A client sends a request: "We want to automate pick-and-place. Universal Robots UR10e or KUKA?" This question can't be answered without numbers from their shop floor. A cobot is an excellent machine for a task that suits it. For a task that doesn't, it's an expensive toy that throttles your takt by 60% against what a classical fenced robot would handle.
First decision point — cycle time
Not safety, not robot price, not integration. **Cycle time on the real task.**
- Cobot on pick-and-place with arbitrary orientation: **8–12 seconds** per cycle. Universal Robots UR5e/UR10e, Doosan M0609, Techman TM5 all sit in the same window. The limit is in the safety-rated monitored stop (SRMS) — TCP speed when an operator is detected must be ≤ 250 mm/s, so accelerations and peak speeds are software-clipped. The cobot reaches full speed only inside a fenced zone, but if you fence it, you've bought a slow robot at the price of an expensive robot.
- Classical fenced robot on the same pick-and-place: **2–3 seconds.** FANUC LR Mate 200iD/7L, ABB IRB 1200, KUKA KR Cybertech KR 6 R900 — all sit around 2 seconds for the same trajectory. At 8 hours a day that's 1,800 cycles vs. 14,400 cycles. Four times more.
If cycle time isn't the bottleneck (the downstream process takes 30 seconds anyway), a cobot is fine. If your line takt is 4 seconds, you cannot use a cobot — not with the best integration. This project is dead before purchase.
Safety integration — the hidden cost block
The arm's sticker price is only 30–50% of the real CAPEX of a robotic cell. The other half is safety.
Cobot — ISO/TS 15066 risk assessment
ISO/TS 15066 (the technical specification for collaborative robots) requires a **biomechanical analysis** — on collision with a human, the force/pressure on the body must not exceed table thresholds (e.g. forehead 130 N, hand 140 N). That means:
- Risk assessment documentation — typically **€3–5k** from a certified integrator or external expert (TÜV SÜD, DEKRA).
- Power-and-force-limiting validation with force-torque calibration — measurement of collision forces on the real trajectory. **€2–4k** including the measurement device (CoboSafe, GTE Pilz).
- Possibly additional sensors — area scanner (SICK nanoScan3, ~€2–3k), safety mat (~€1–2k), light curtain at zone entry (~€1.5–3k) depending on the SRP/CS level.
**Total €5–10k of safety infrastructure for a cobot.**
Industrial robot — full fence
ISO 10218-2 (industrial robots) requires physical separation from the operator. Real prices:
- Mesh fence or polycarbonate panels: 30–80 m² × €120–200/m² = **€3.5–16k**.
- Safety doors with interlock (Pilz PSEN ml1p, Sick i110, Schmersal AZM 300): 2× €1.5–3k = **€3–6k**.
- Light curtain or laser scanner at entry (SICK miniTwin, deTec4 Core): **€2–5k**.
- Safety PLC (Pilz PNOZmulti 2, Siemens S7-1500F): **€3–5k**.
- Wiring + category 3/PL d certification + documentation under EN ISO 13849-1: **€4–8k.**
**Total €15–30k of safety infrastructure for a fenced robot.**
The cobot therefore saves €10–25k on safety. But loses it to the four-fold cycle time. The crossover is decided on production per shift.
Payback — two real examples
Example 1: Small electronics batches, single-shift operation
- Product: test + pack an electronic board, ~800 units per shift.
- Manual: 1 operator, hourly cost together €18/h (wage + contributions + overhead).
- Takt: 30 seconds per piece (not the bottleneck — the test stand takes 25 s anyway).
- Investment cobot UR5e + gripper + vision + safety: **€55–65k**.
- Investment fenced robot ABB IRB 1200 + fence + safety: **€70–85k.**
At 1 shift × 5 days × 50 weeks = 2,000 h per year × €18 = **€36k/year wage savings.**
Cobot ROI: 65/36 = **1.8 years.** Fenced robot ROI: 80/36 = **2.2 years.**
Here the cobot wins — not just because of 15–20% lower CAPEX, but also because a small batch means frequent reprogramming and product changes. A cobot reprograms in 2 hours by manual guidance (kinaesthetic teaching) + UR PolyScope. A fenced robot in RobotStudio / KAREL / RAPID takes 1–2 days with offline simulation. At one product change a month, that's 12 days of engineering work per year extra — €8–12k. The cobot wins twice.
Example 2: Tier-2 automotive supplier, 3-shift operation
- Product: welding + machine tending, **6,000 pieces per shift, 3 shifts daily.**
- Takt: 4.5 seconds per piece. This is definitively outside the cobot window.
- Manual: 2 operators per shift (one welding, one material handling) × 3 shifts = 6 FTE × €22/h (auto industry, more overhead + night premiums) × 2,000 h = **€264k/year.**
- Investment fenced robot FANUC R-2000iC/165F + fence + safety + welding head: **€140–180k.**
- Investment cobot — **impossible, the takt is unreachable.** Even if you put two cobots on a parallel process, you're still under 50% of the fenced robot's output with comparable CAPEX.
Fenced robot ROI: 160/264 = **0.6 years.** Seven months and the machine is paid off, 9.5 years of operation go straight to the P&L.
In Tier-2 automotive a cobot in this profile **is never used.** Not for ideology — for takt arithmetic.
Cobot without hesitation when
- Small batches (< 1,000 units/day), frequent reprogramming, mixed products.
- Shared workspace with a human (quality control, loading complex parts, where the human adds visual inspection).
- Cycle time > 8 seconds that can't be shortened because of another node in the line.
- A startup operation without anyone certified for industrial robots. Cobot programming is something a mechatronic technician with 1 month of training can handle.
- Office-like environments (laboratory, pharmacy dispensary, low-volume electronics) — where a fence looks brutal and ruins the workplace aesthetic.
Fenced industrial robot without hesitation when
- Takt < 5 seconds per cycle.
- 2+ shift operation, where the fence amortises quickly.
- High payloads (> 15 kg) — cobots do exist here (Doosan H2017, FANUC CRX-25iA), but at a price that wipes the advantage and with even slower takt.
- Welding, grinding, dispensing with reactive media — safety risks that ISO/TS 15066 power-limit doesn't cover. Sparks from MIG/MAG welding a cobot can handle, but an operator in a 1-metre zone during welding is an OHS problem nobody will sign off.
- Existing robot fleet where the team speaks RAPID/KAREL/KRL — adding another cobot with its own ecosystem is organisational friction.
Hybrid setup — the most common real choice for mid-sized firms
In a fully automated line for volumes of 2,000–4,000 units per shift, the typical mix is:
1. **Main cycle = fenced robot.** Fast handling, high payload, operator removed from the zone. 2. **Side station with a cobot.** Visual inspection, loading material onto the line input, replacing a cassette that doesn't pay back to fully automate. 3. **Shared safety architecture.** One Pilz PNOZmulti 2 or Siemens F-CPU handles both the fence around the IR and the power-and-force-limit controllers for the cobot. Shared HMI, shared maintenance.
This split typically lands at 60% of CAPEX for the fenced robot and 40% for the cobot — and delivers the line's full takt without compromise in zones where a human has to be anyway.
What to check before PO
Regardless of choice, four checks before PO:
1. **Real payload + speed combo.** A UR10e with a 12.5 kg payload is NOT the same as with 3 kg. At max payload, speed drops 30–50%. The datasheet shows both separately; measurement shows the real number. 2. **Reach + singularity zones.** Check kinematics in simulation (RoboDK, Process Simulate, Universal Robots Studio) for your real trajectory — not a trade-show demo. 3. **Cycle time test on your part, not on a Lego brick.** The distributor will gladly show you a demo. Ask for cycle time on your part, with your gripper, on your trajectory. The difference is usually 30–100%. 4. **Total cost** = robot + gripper + safety + integration + offline programming software (RobotStudio €4–8k, RoboDK €1.5k licence, KUKA.Sim €5–12k) + team training (3–5 days × €1.5k/day). The real sticker is usually 2.2–2.8× the arm price.
---
*We write this as a technical partner that has integrated both types in the same client environments over recent years. If a specific use case interests you, the first consultation (90 minutes) walks through takt, payload and safety profile on your actual task and gives you an indicative CAPEX before anyone sends you a quote.*