Robot integration cost, line by line.

The most quoted version of this rule comes from Universal Robots' co-founder Esben Østergaard, who told Machine Design in 2018 that with traditional robots the arm accounts for only 25 to 30 percent of total system cost. Flip that around and the cell costs 3.3 to 4 times the robot. He was making a case for cobots, so take the exact number with a grain of salt. But it matches what we see on real quotes, and it matches our own industrial robot cost guide, which puts a working cell at two to four times the arm.

The multiplier moves with how much of the cell is custom. A packaged cobot palletizer reuses the same frame, software and safety layout on every install, so the integrator's hours are low. A vision-guided bin picking cell needs 3D camera work, path tuning and a lot of testing on your parts. Multi-station assembly is the extreme case, because every station is its own little machine.

Treat the multiplier as a sanity check, not a pricing method. A $28,000 cobot with a $60,000 vision system gives a wild ratio and can still be a fair quote. Price the cell from its parts, as the rest of this page does, and use the ratio to spot outliers.

FAT and SAT, the two lines people cut

A factory acceptance test means the cell runs your real parts at the integrator's shop before it ships, against criteria you agreed in writing: cycle time, pass rate over a set run, recovery from a dropped part, every safety stop. A site acceptance test repeats the important parts on your floor after install. Together they're maybe 6 to 10 percent of the hours on our example.

They're also where problems get cheap. A gripper that slips on oily parts costs an afternoon to fix at the integrator. The same fix on your floor, with a lathe waiting and an operator standing there, costs days. When a quote skips FAT to look lean, that debugging still happens. It just happens on your production schedule.

Safety is a line, not a footnote

The cell has to be designed to the current robot safety standards, which in the U.S. now means ISO 10218-2:2025 and ANSI/A3 R15.06-2025. That starts with a written risk assessment for your specific cell and ends with guarding, scanners or speed-and-separation settings that match it. If a quote lists guarding but no risk assessment, ask who's doing it. Usually the answer is "you," and most plants aren't set up to.

Our qualification and safety work covers what that assessment includes and what you should keep on file after handover.

An hourly rate looks high until you build it from the paycheck up. The Bureau of Labor Statistics puts the median mechanical engineer at $104,110 a year in May 2025, and electrical engineers at $120,630. Add payroll taxes and benefits. Then remember that nobody bills 2,080 hours a year. Proposals, training, travel and the job that slipped all eat into billable time.

After that comes overhead the integrator carries whether or not you sign: the shop floor where your cell gets built, CAD and offline programming licenses, demo robots, insurance and the sales engineer who scoped your project. The math below gets you to about $130 an hour before anyone's made a profit, which is why a $95 rate should make you curious rather than happy.

Rates also hide in plain sight. Some integrators quote a lump sum and never show hours. Others quote a lower rate with a higher hour count. Ask for hours by phase either way. It's the only way to tell whether a $40,000 programming line is 250 hours of careful work or 120 hours at a premium.

Think about what the arm doesn't know. It doesn't know where your part is, how to hold it, what to do when it drops one, or how to tell the lathe it's safe to close the door. Every one of those is someone's afternoon, and on a high-mix, low-volume floor it's someone's afternoon per part family.

Then there's the stuff around the robot. Fixtures get machined to your part, guarding has to fit your aisle and forklift route, and the controls panel has to talk to a machine that may be twenty years old.

Recovery logic is the part that surprises people. The happy path of a pick-and-place program can be written in a day. Getting the cell to handle the ten ways a cycle goes wrong, without an operator walking over every time, is where half the programming hours go. It's also what separates a cell that runs a second shift alone from one that needs a babysitter.

Robots, safety scanners and some cameras can carry lead times of several weeks, so procurement starts as soon as the design freezes. That's also why design changes late in the project cost more than they look. A changed gripper isn't just new fingers. It can mean a new reach study, a re-ordered part and a slipped FAT.

Watch the payment terms as closely as the price. A quote asking for 50 percent up front and the balance on shipment leaves you with no money held against a cell that hasn't run on your floor. We'd rather see the big payment tied to a FAT you witnessed, and a real slice held until SAT.

Fixed price only protects you if the scope is written down. That means the part list with revisions, the cycle time, the acceptance criteria and what counts as a change. Without those, a fixed-price job turns into a stream of change orders, and change orders are priced at time and materials anyway.

Time and materials makes sense in two places. One is a second or third copy of a cell you already own, where the unknowns are gone. The other is a short, paid feasibility study before the real quote: a few days of testing your parts with a gripper and a camera to settle the scary questions. We like that second use a lot. It turns the biggest unknowns into facts before anyone signs a six-figure number.

What a hybrid looks like

A common middle ground is a fixed price for the defined scope, a published rate card for changes, and a cap on travel and expenses. The rate card matters more than people think. It's what you'll pay for every "can it also do this?" after go-live.

Red flags in a robot integration quote

• Engineering and programming as one lump with no hours.
• No FAT, or a FAT "with sample parts" instead of yours.
• Guarding listed, risk assessment missing.
• Installation, rigging or electrical hookup "by customer."
• Cycle time promised with no assumptions written next to it.
• No acceptance criteria, so nobody can say when you're done.
• Over 50 percent due before the cell has run your parts.
• No rate card for changes, or one that's much higher than the base rate.
• Software, source code or program backups not delivered to you.
• Warranty that starts at shipment instead of at SAT.

None of these makes a quote dishonest. Good integrators write lean quotes because buyers reward the lowest number. But each one moves a cost or a risk onto you.

The acceptance criteria deserve special attention. "Cell will tend the lathe" isn't a criterion. "Cell runs 500 consecutive parts of P/N 1234 rev C at 62 seconds or less, with no more than one operator intervention" is. If you can't write the sentence, the integrator can't price it, and the fight about whether the job is finished starts on install day.

If you're choosing between integrators rather than quotes, our ranking of robot integrators for high-mix manufacturing explains the options, including when a turnkey cell beats a custom one.

Fix the part presentation first. Most of the cost difference between an easy cell and a hard one is how parts arrive. Parts in a known orientation on a tray or a fixture need a simple gripper and a short program. Parts dumped in a bin need 3D vision and a lot more testing, which is fine when it pays, but it shouldn't be the default. Often a $2,000 tray change on the upstream process saves $30,000 in the cell.

Next, reuse beats invention. A cell built on a proven base (a standard frame, a standard safety layout, a program template for machine tending) carries far fewer design hours than a blank sheet. That's the logic behind packaged cells and behind our own Relling workcell. And buy the robot the integrator knows best unless you have a plant standard. Asking a FANUC shop to program its first ABB arm, or the reverse, adds hours you'll pay for. The FANUC vs ABB comparison covers what actually differs between the two.

Be careful where you cut. Dropping the FAT, skipping the risk assessment or doing your own install rarely saves money overall. Each one moves work to the most expensive place to do it, your own floor.

One saving people miss is the second cell. The second copy of a cell often costs well under the first, because the design, programming and testing are already paid for. If you know you'll want three, say so in the first RFQ and ask for the price of each.

Finally, check that the cell pays back before you squeeze the price. A cell that's 10 percent cheaper and runs 20 percent fewer good parts is a worse deal. The robot ROI calculator runs that math, and the robot cell cost breakdown goes deeper on the hardware lines.