The brief
A collaborative-robotics startup had shipped their first 50 units using prototype-priced parts machined by a shop that did great one-off work and had no intention of scaling. They came to us with their joint housing drawing, the 50-unit cost they were paying, and a target of 4,000 units across six months at a unit price that was 30 % lower. The housing is the structural body of one of their arm joints: it holds a bearing, mounts to a servo flange, and bolts to the next link in the arm.
First conversation — the DFM pass
We did not quote on the prototype drawing. We asked for a 30-minute call with their mechanical lead and walked through seven features where their drawing was over-specified for what the part actually needed. Some of these were historic ghosts from their prototype shop’s default settings; some were conservative choices the mechanical lead had never gone back to review. All seven came off the drawing in a single revision cycle. That one call cut cycle time by almost 40 %.
This is the step that separates a first-unit shop from a production-scale shop. At 50 units, the DFM round isn’t worth the time — just machine the drawing. At 4,000 units, a 40 % cycle-time reduction is $18,000 of machine time over the run. We do not charge for the DFM pass on jobs we quote.
Fixture and process
We designed a pallet fixture carrying four parts per cycle on our DMG Mori NMV5000 cell. First operation machines five faces from billet, including the bearing bore to 0.05 mm oversize. Parts are flipped and re-clamped for the second operation, which finishes the remaining face and brings the bearing bore to final dimension with a single-flute reamer. Total cycle time at lot 8: 11 minutes per part including load/unload.
We keep the bearing bore finish pass as a separate pass with a fresh reamer every 200 parts, indexed by tool-life counter. This is the feature that has to hold ±0.015 mm and it is worth the consumable cost to keep the reamer sharp. We’ve seen shops try to run thousands of parts off one reamer and then scrap a whole lot when the drift catches up.
Anodize qualification
We run Type II Class 2 black anodize on this part — about 25 μm thick, dyed black, sealed with hot DI water. The customer specified Delta-E ≤ 2.0 across shipments, which is achievable with a well-controlled line but not automatic. We qualified a local Dongguan vendor with a 200-piece pilot, measured coating thickness in five locations on ten parts, ran salt-spray per ASTM B117 to 336 hours, and compared color to a master swatch under D65 lighting.
Color drift is the thing that bites you on black anodize over long runs. Our protocol: ship 10 pre-anodize parts to the vendor as a color-match calibration before each shipment of 500, re-run the calibration if anything changes in their dye bath. We caught one bronze-toned batch at incoming inspection, had the vendor re-anodize on 48-hour turnaround, and the customer never saw a mismatched shipment.
Cost reduction over the ramp
Lot 1 (December 2025): unit cost baseline. Lot 8 (April 2026): $1.12 per unit cheaper. That reduction shipped through as customer price breaks at 500 units and again at 2,000 units — we did not keep the savings, we passed them through in exchange for volume commitment. That is the partnership shape most of our robotics customers want and it is the one that rewards both sides when the volume actually materializes.
Where this model works
Parts in the 500-to-10,000-unit range, with at least one feature that needs real CNC tolerance, and a customer willing to iterate on DFM. If you’re at that volume band and your current vendor is treating you like a prototype shop, the DFM pass is free — send us the drawing and we’ll tell you honestly whether there are savings to find.