Our published customer studies are still coming. Until they're ready, we publish representative deployments: the cell we'd build for a common shop profile, run through public wage and inspection data, with every number tagged so you can see which ones we measured (none), which we looked up, and which we chose.
One so far, a representative robotic deburring case study. It covers the cell layout, a cycle timing diagram, baseline vs result, payback worked step by step, sensitivities, and what has to be true at your plant for the numbers to transfer.
A high-mix machine shop with 40 part numbers and a hard-to-staff second shift. A cobot with a compliant spindle and a smart camera, modeled at 1.46 operator minutes a part and about 19 months to payback.
Every figure in a representative study carries one of four tags, shown inline next to the number.
There's no "measured" tag yet. That one is reserved for published customer studies.
It's a worked example of a real cell design applied to a shop profile we see often, not a story about a particular plant. Nothing in it was measured, so every number shows where it came from, and the payback is computed in the open so you can redo it with your own figures.
A shop type that keeps showing up in scoping calls, with its volumes and pain points written down as assumptions.
Stated, not sourcedRobot, tooling, vision, guarding and part presentation, with specs from vendor datasheets and planning prices.
Inside the $100k to $200k bandWages from BLS, inspection performance from published research, process data from NIST and vendor sources.
Every source linkedCompute results from the inputs only, show each step, then move one input at a time to see what breaks the case.
Sensitivity on every studyWhy publish these at all? Because the questions buyers ask before a cell, like what it costs, how long a part takes, what it does to labor and when it pays back, have answers that depend mostly on inputs you can check. A study with its inputs showing is more useful to you than a glossy before-and-after you can't test. We'd rather show you the labels than hide them.
If you want to run the same math on your own cell, the robot ROI calculator takes the same inputs, and the robot cell cost breakdown explains what sits behind each line of an installed price.
Plant data from a stated measurement window, the customer's sign-off, and the same tagging, with measured results marked as measured. We'll publish them once cells have run long enough to measure and customers agree to be named.
| Representative deployment | Published customer study | |
|---|---|---|
| Plant | A composite shop profile, unnamed | A real plant, named with permission |
| Baseline | Stated assumptions plus public benchmarks | Timed and logged on the floor before install |
| Results | Modeled from the inputs, math shown | Measured over a stated window, with the method |
| People and quotes | None | Only with the person's approval |
| What it's good for | Learning the method; first-pass budgeting | Evidence that a similar cell worked somewhere |
Table scrolls horizontally on small screens →
Want to see how a cell like these would look on your own parts? The Relling workcell page covers the platform, and the finishing and deburring page covers the application in the study above. For the deeper technical side, start with burr and edge inspection with machine vision or how to automate deburring.
Relling builds turnkey robotic workcells for high-mix American manufacturers, using arms from FANUC, ABB, Universal Robots, Yaskawa, KUKA and others, and we're paid by none of them. Send drawings and a week of timings and we'll run the same model you see in these studies on your parts, including the cases where the answer is "not yet."
Talk to an engineer →We started Relling to help American manufacturers make more of what this country needs. Tell us about the part and we'll scope a cell to it.