Robot cell or a second-shift operator? How to work out the payback
A robot cell pays back through spindle hours: the nights a machine stands still today and the loading time it takes off your operator. In our worked example, a one-shift machine and an assumed €120,000 cell (a round number, not a SIMUL price) pay back in about 28 months, slow start included, and add about €151,500 over five years in today's money. Whether a cell, a second-shift operator or both makes most sense for you depends mostly on how much repeat work you can sell and how many of the planned night hours actually run, and the calculator below shows where the answer changes.
Your machine probably cuts less than you think
Ask a shop owner how busy the machines are and the answer is usually 60% or more. Bill Bither, co-founder of the US machine-monitoring company MachineMetrics, said in a 2020 webinar reported by Modern Machine Shop that this is what most users assume before they install its monitoring system, and that the measured average is generally around 25%. That is a vendor figure from its own customers' machines, counted against all hours of the day, but European data points the same way. In Tooling in Germany 2020, a study by the WZL of RWTH Aachen University and Fraunhofer IPT, German tool and mold shops showed about 92% utilization on their milling machines, measured against the hours their staff are present. In the same study, the milling machines of the tool shops in its 2019 benchmarking competition ran 2,835 hours a year on average: 32% of the 8,760 hours in a year.
Both numbers can be right, because utilization is a ratio and the denominator decides it. Modern Machine Shop's Top Shops survey uses the narrowest one, in-cut time as a share of machine uptime: in its first round, covering 2010, the best 20% of 183 shops reported 80% and the others 60%. Take one machine that cuts 1,104 hours a year, and it can truthfully report three different numbers.
For a robot cell only the last bar matters, because the hours a cell can add are the hours outside today's cut. Where the stops come from matters too. When Chalmers researchers analyzed six months of automatic stop logs from 884 machines in 23 Swedish companies in 2013 and 2014, machine shops among them, about 90% of the stop time they could classify was tied to support work done by operators, such as changeovers, measuring and adjustments, and not to the machines' own process.
So measure your own machine for two normal weeks before anything else. Many controls keep power-on, operating and cutting-time counters; a monitoring system or a paper log at the machine also works. Write down why the spindle stopped, then sort the stops into two groups: those a robot removes (loading and unloading, waiting for the operator, the hours nobody is there) and those it does not (setups, missing programs, missing material). Only the first group belongs in the payback.
Where the extra hours come from
How many of the 8,760 hours a machine can cut depends on how you staff it. The example uses 230 working days a year (365 days minus 104 weekend days and 31 days of holidays and vacation).
| Pattern | Arithmetic | Hours a year | Share of 8,760 |
|---|---|---|---|
| One shift | 230 × 8 h | 1,840 | 21% |
| Two shifts | 230 × 16 h | 3,680 | 42% |
| Around the clock on workdays | 230 × 24 h | 5,520 | 63% |
| Calendar year | 365 × 24 h | 8,760 | 100% |
A one-shift machine cutting 60% of its scheduled hours cuts 1,104 hours a year (0.60 × 1,840), 12.6% of the calendar. The gap between 1,104 and 8,760 hours is where a robot cell finds its hours, and where a second shift would find them too.
Robot cell or a second-shift operator?
Both add hours. They cost and deliver them differently.
An operator is the flexible option. Covering a second shift of 1,840 hours costs about €101,200 a year at Denmark's average labor cost in manufacturing, €55.0 per hour worked in 2025 by Eurostat's figures, before the shift premium in your collective agreement. At Germany's €49.5 it is about €91,100. If that operator runs only this machine at 60% in cut, each hour in the cut costs about €92 in labor alone (€101,200 ÷ 1,104 hours); shared over two machines, about €46. In return, the operator handles new parts, first-off inspection, changeovers and the alarm at 22:00, which a cell does not.
A cell is the cheaper hour once it runs. In the worked example it costs about €23 per added hour in the cut: the €120,000 price less a 25% resale value after five years, spread over those five years, plus €20,965 a year in running cost, which includes the day-shift time to stock and restart the cell, divided over 1,665 added hours. Those hours exist only if the day shift prepares repeat work for the night, and preparing new work takes operator time. Maskinfabrikken HMA, a Danish subcontractor that went from one robot arm to four, gave Dansk Industri a hypothetical example in 2023: an operator spends seven hours getting the robot ready to make 10 parts that take 10 hours the traditional way, and the shop gains three hours. HMA also had to hire a couple of new people because its robots raised output. "The robot arms are 100% dependent on us," CEO John Richter told DI (our translation).
The cheaper hour does not settle it. The cell can only fill its nights with repeat work, its fixed costs need enough sellable hours to carry them, and an operator who runs several machines costs each of them less. The calculator puts all of that side by side.
Run the numbers for your shop
The calculator compares three ways to add spindle hours to one machine against changing nothing: a robot cell on today's shift, a second-shift operator, and both together. Every value is an input. The example values are ours, chosen to be plausible for a small European shop, and the €120,000 cell price is a round number for the example, not a price for any specific cell. All results are for this one machine. The calculator runs in your browser and sends nothing anywhere.
The worked example, step by step
With the example values, the cell adds 1,665 hours in the cut a year. In the staffed shift the robot loads while the operator does other work, and in-cut time rises from 60% to 70%: 184 hours. At night it runs a 12-hour window on 230 working nights. A storage load of 40 parts at 20 minutes, plus a 30-second swap per part, lasts 13.7 hours, so it covers the night; the spindle cuts 97.6% of the running time; and we assume 55% of the planned night hours actually run, after alarms, broken tools and nights with nothing loaded. That is 6.4 hours a night, 1,481 hours a year.
At €50 an hour (an €80 hour rate minus €30 that rises with each hour in the cut), the hours are worth about €83,200 a year. Running cost is €20,965 a year: €3,000 in service, €9,000 for gripper fingers and fixtures for 6 new part numbers, €2,640 for 8 hours of programming and proving each, and €6,325 for half an hour a day of day-shift time to stock and restart the cell, all labor at €55 an hour. The gain at full rate is about €62,300 a year. With the first six months at half rate, the cell pays back in 28 months, and over five years it adds about €151,500 in today's money (net present value at 6%, counting a 25% resale value at the end).
A second-shift operator who runs this machine and one more adds 1,104 hours for €50,600 a year, this machine's half of the labor cost, plus half of €5,000 for hiring and training, and adds about €10,600 over five years. Both together add 2,459 hours, with a shorter 8-hour night between two shifts, and about €98,800.
The answer changes when the inputs do. With everything else at the example values, the second-shift operator comes out ahead if fewer than about 29% of the planned night hours actually run, if 23 or more new part numbers a year go onto the cell, or if the cell costs more than about €293,000. The cell and a second shift together come out ahead if less than about 48% of the extra work is repeat work, if the operator runs three machines, if the hour rate is above about €96 or labor costs less than about €42 an hour, or if the cycle time is under about 14 minutes. Below 17.5 minutes one storage load no longer lasts the 12-hour night, while it still covers the 8-hour night between two shifts. If the order book takes less than about 1,000 extra hours a year, or the hour rate is below about €58, none of the three pays within five years.
This is a model of one machine, and it leaves out tax, financing terms and the value of shorter lead times. For a board decision, run the same cash flows with your accountant.
Routine first, then the robot
A large public data set on unattended machining and profit comes from the Top Shops surveys by Gardner Intelligence, publisher of the US magazine Modern Machine Shop. In an analysis presented in 2019, about 1,100 shops said whether they use a formal improvement methodology. Those without one had a median profit margin of about 1%, and the few of them that ran unattended were no better off: they stayed at 0 to 1%. Among shops with a methodology, unattended running went with a margin of 8% against the group's median of 5%, and 9% when they also monitored their machines.
The data is self-reported and correlational, but the mechanism is easy to see at the machine. A night run with unproven programs, unplanned work and tools of unknown life stops at the first alarm, and the night hours you planned never arrive. Sweden shows the same pattern: a 2019 Chalmers study of three sheet-metal subcontractors found their robots, which tended press brakes and a press, were used very little, because under delivery pressure nobody had time to program the robot or design grippers. One of the shops judged that a job needs at least an hour of machine time to be worth automating.
The same surveys show that unattended hours do not always need a robot. In the 2018 Top Shops round, answered by 471 shops, 59% of the best 20% ran machines unattended, while 23% used machine-tending robots. Bar feeders, multi-part fixtures and pallet pools were the other ways to run unattended that Modern Machine Shop named.
When a cell does not pay
You cannot sell the extra hours. A recovered hour with no order behind it earns nothing, and the calculator shows what a small order book does to payback.
The work does not repeat. Every new part number needs its own gripper fingers, fixture, program and a proven first night, and that engineering returns only if the part comes back. In a 2015 survey of 825 Danish manufacturers with 5 to 250 employees by the Danish Technological Institute, firms that used no robots most often said robots did not suit their process or that their production was too small; remarkably few named the price.
Short cycles. Storage empties fast, and loading time becomes a large share of the cycle. If the robot needs 30 seconds to swap parts on a 2-minute cycle, the spindle stands for 30 of every 150 seconds, 20% of the time.
Parts the gripper cannot handle: too heavy once the gripper, and a pallet or vise if the robot carries one, are added; no repeatable gripping surface; finished surfaces that cannot take a gripper finger; raw stock that varies too much to locate.
Processes that need a person mid-run: a deburring or inspection step between operations, or chips that pack into deep pockets and need clearing.
When a cell pays
Repeat parts with proven programs and enough volume to fill the unattended window. A night that runs a new program for the first time is a night that stops at the first alarm.
Cycle times long enough for the part storage to cover the night. Autonomy is parts in storage × (cycle time + swap time): 40 parts at 20 minutes plus a 30-second swap is 820 minutes, 13.7 hours, enough for a 12-hour window. The same 40 parts at 3 minutes last 2.3 hours.
Parts finished in one clamping. One load and one unload per part is the simplest job a robot can have, which is one reason 5-axis and automation work well together. Our article on whether your shop will benefit from 5-axis covers the clamping count.
Standard workholding. A zero-point pallet system gives the robot one gripping interface for many parts, so a new part number needs a new fixture on a pallet instead of a new gripper.
What keeps a machine cutting unattended
The share of the night window in the cut depends on process security more than on the robot. Tools wear, so sister tools take over: the Unicorn 5-200 Modular holds 78+1 tools. Tools break, so breakage detection between cycles matters: the optional Renishaw NC4+ Blue F145 tool setter does that. Chips have to leave the work area (an optional chip conveyor), and workholding has to be machine-controlled (an optional hydraulic feed-through machine table with an AMF 6204IHA K10.2 zero-point interface, and an optional zero-point vise pallet with a Spreitzer HZS 120-60 A hydraulic centric vise and a robot gripping interface).
SIMUL's robot cell uses a long-arm FANUC LR Mate with a 14 kg payload, gripper included, and is designed to dock to the Unicorn 5-200 Modular through its CNC-controlled 410 × 308 mm automation door, which every machine has. Every machine is made ready for robot machine tending, so the cell is designed to be added when your volumes call for it. The cell is quoted per configuration, because the gripper, the part storage and the cycle logic follow your parts. The automation page shows how it works, and Options and Pricing shows how each part is priced.