The first question a buyer usually asks us is which brand of robot arm to buy. It is the wrong first question, but it is a fair one, and we answer it — after we have asked for the part drawing, the cycle time target and a photo of the floor space.
Here is the pattern we see. A factory decides to automate one station, gets three quotes, and compares the arm prices line by line. The arm is maybe 30–45% of the total. Grippers, fixtures, guarding, integration, safety hardware and commissioning make up the rest, and those costs are not proportional to the arm price at all — a cheap arm in a badly planned cell costs more than a premium arm in a good one. So we start with the cell, not the arm.
Start with the task, not the robot
Write down what the operator does now, in order, with times. Not the job description — the actual sequence: pick from a bin, orient, place in a fixture, press a pedal, wait eleven seconds for the machine, unload, stack, repeat. If you cannot describe the cycle in that detail, no robot quote will be meaningful, and any supplier who gives you a firm price without asking for it is guessing.
Once the cycle is written down, three numbers drive the whole specification.
Payload, honestly measured. Payload is not the weight of the part. It is the part plus the gripper plus the adapter plate plus any hose or cable bundle dragging on the wrist. A 5 kg part with a 3 kg magnetic gripper and a 1.5 kg adapter needs a robot rated well above 9.5 kg, because rated payload is measured at a specific centre of gravity, usually close to the wrist flange. Move that mass 200 mm out and the usable payload drops — sometimes by a third. Manufacturers publish the payload-vs-offset curve for a reason; ask for it before you accept "6 kg is plenty."
Reach, measured, not estimated. Published reach is to the wrist centre, from the mounting flange, with no tool. The point that actually matters is where your gripper fingers need to be, in the worst-case position of the cycle. Sketch the cell from above, mark the pick point, the place point and the tool-body working envelope, then draw the reach circle. If the far corner of the fixture sits outside it, the operator will spend the next five years hand-loading that corner. On articulated 6-axis arms in the small-to-mid class, reach commonly runs from roughly 700 mm up to about 1,700 mm; each step up costs money and often requires a bigger pedestal or a longer guard fence.
Cycle time with margin. Take the robot's rated cycle time — usually quoted for a light payload over a short path, in ideal conditions — and add real-world losses: gripper open/close time, air pressure recovery, position-system waits, vision inspection, operator intervention. A rule we use is to budget the robot at no more than 70–75% of the target cycle, so a late change in gripper weight or a slower machine tool does not sink the business case. This is usually where an automation project stops being a good idea, and it is much cheaper to find that on paper than on the shop floor.
Cell design decisions that decide the cost
Mounting. Floor-mounted is cheapest and simplest. Overhead gantry mounting buys reach over a machine but adds structure, alignment work and money. Wall mounting saves floor space but loads the column differently. Choose floor mounting unless you can explain why the ceiling is the right answer.
Mounting surface. Robots are bolted to whatever you have. A 10 mm steel plate on a reinforced slab is fine. A mezzanine or a raised platform with a wooden core is not, and the vibration shows up as poor repeatability long before the robot fails. If the floor is a slab, say so; if it is not, budget a foundation.
Guarding and safety. A conventional industrial arm needs a fence, an interlocked gate and light curtain or scanner, plus a safety-rated stop circuit. This is not optional and it is not free — for a small cell, guarding and safety hardware typically add a few thousand dollars and can add considerably more if the cell sits in a busy aisle where operators pass. Collaborative robots reduce the fence requirement for some applications, but they do not remove it: outside force and speed limits apply, pinch points still exist, and a cobot running at 1.5 m/s next to a human is a different risk category from a cobot running slowly with force monitoring. Buy the cobot for the flexibility and ease of redeployment, not because you think it exempts you from a risk assessment.
Positioning system. A robot is only as accurate as the positioner feeding it. A quality servo positioner with an encoder interface costs real money; a cheap indexing table will limit the cell to whatever tolerance it can hold, forever. If your part needs a two-degree rotation between operations, plan for it in the fixture rather than in a second robot.
Services. Compressed air, electricity and network. Air is the one that catches people: most grippers are pneumatic, and a cell running a fast open/close cycle will draw far more air than the shop line was sized for. Check pressure at the point of use, not at the compressor. A small dedicated receiver tank near the cell is an inexpensive fix that prevents a lot of cycle-time mystery.
| Cell element | Typical share of project cost | Where buyers get burned |
|---|---|---|
| Robot arm | 30 – 45% | Chosen before the cycle is defined; payload counted without the gripper |
| End effector / gripper | 5 – 15% | Custom gripper designed last, holding up the whole project |
| Fixtures & positioners | 10 – 20% | Tolerance forgotten; indexing table too coarse for the part |
| Guarding & safety hardware | 8 – 15% | Budgets it as "fencing" and ignores light curtains and interlocks |
| Integration, programming, commissioning | 15 – 25% | Quoted as a lump; nobody owns the risk/benefit analysis |
| Spares & training | 3 – 8% | Cuts training to save money, then runs the cell at half speed |
Rule of thumb from our own quotes: a complete single-station robot cell lands at roughly 1.8× to 3× the price of the bare arm, depending on gripper complexity, guarding and integration. If a quote comes in at 1.1× the arm price, read what is missing. If it comes in at 5× without explanation, read what is padded.
The numbers that decide payback
Before we recommend anything, we run a two-line payback calculation with the buyer. It is not sophisticated, and that is the point.
Line one: labour replaced. How many operators does the station consume today, and how many after automation? Be conservative. One operator per shift, three shifts, plus a fraction of a supervisor's time is a normal honest figure; assuming the cell runs unattended through breaks and holidays is not.
Line two: what the cell costs to run. Electricity, compressed air, maintenance, spares, programming time for every product change, and the depreciation you decided on. Then add availability: an industrial robot in a well-maintained cell typically runs 90%+ of scheduled time, but that assumes someone in the plant can program and recover it. If nobody can, the real availability is much lower, and the payback moves accordingly.
Where the arithmetic works, it usually works because the cell runs multiple shifts on a stable product mix. Where it fails, it fails because the product mix changes every few weeks and the cell has to be reprogrammed for each change — that cost is real, and it does not show up in the robot quote.
Mistakes that cost buyers the most
- Buying a cobot because it sounds easier. Cobots trade speed and stiffness for flexibility. On a task that runs the same way for three years, an industrial arm is often cheaper and faster. On a task that moves between products every week, the cobot wins.
- Underestimating the gripper. End effectors are usually the longest lead item in the project and the most likely to need a second design. Get the gripper concept agreed before you sign the robot PO.
- Forgetting the operator. Somebody has to load the magazine, clear jams, inspect the first part and stop the cell. Design that station's position and access route at the same time as the robot's.
- Treating integration as a line item. It is the part of the project most likely to run over, and the supplier with the cheapest integration quote usually has the least defined scope.
- Ignoring the spare parts list at signature. Order the spare gripper fingers, the teach pendant battery, one spare servo drive and a set of fuses with the cell. Getting them six weeks into a production ramp is expensive in a way the price list never shows.
The blunt version: define the cycle, measure the payload with the tool, measure the reach to the worst-case point, budget the robot at 70–75% of the target cycle, and plan the gripper, guarding and positioner before you compare arm prices. Do that and the robot choice mostly makes itself. Skip it and you will buy a perfectly good arm for a cell that cannot run.
If you are planning a robot cell and want a second opinion on the specification, send us the cycle description, the part drawing and a floor sketch. We will tell you which class of arm fits, what the cell will realistically cost, and where the payback calculation is most likely to break.