By the time a component reaches a spindle, its cost is largely settled, not by negotiation, not by machine capacity, but by decisions frozen into the drawing months earlier. Read one the way a shop does and the number stops looking arbitrary.
The title block, and what the material costs you. Aluminium is forgiving. 15-5PH brings insert life into the conversation. Inconel 718 or Ti-6Al-4V changes the character of the job entirely: surface speeds fall dramatically, the alloy work-hardens the moment a tool dwells, and titanium's poor thermal conductivity drives heat into the insert rather than letting the chip carry it away. Tooling stops being overhead and becomes a priced line. Quietly alongside the alloy sits stock utilisation: a part milled from solid with a heavy buy-to-fly ratio means much of the material bill leaves the building as swarf.
The tolerance block, and which process route each feature takes. This is where cost stops climbing and starts jumping, because tightening a tolerance doesn't make a machine try harder. It changes the process:
| Tolerance | Process route |
|---|---|
| ±0.1 | Ordinary machining |
| ±0.05 | A dedicated finishing pass at reduced depth of cut and feed |
| ±0.02 | In-process gauging, offset correction between pieces, a thermally settled machine |
| Below that | Off the mill entirely, onto grinding |
Steps, not a slope. Which is why relaxing a callout by a hundredth of a millimetre sometimes changes nothing and occasionally deletes a whole operation. Two things hide in this block. Surface finish moves on its own axis, so an Ra specification can force a finishing route with no dimensional justification behind it. And gauge capability sits underneath all of it: a tolerance tight enough to demand very fine instrument resolution and a temperature-stable inspection room carries cost that has nothing to do with cutting metal.
The geometry, which translates almost directly into setups. Every re-fixturing adds handling time and a fresh datum transfer, so a single feature that forces a fourth operation costs more than its size suggests. Beyond setup count, most geometric cost is really a question of rigidity, what the part and the tool will do under load:
| On the drawing | On the floor |
|---|---|
| Datum referenced across the part | Single-setup machining, or a fixture built to preserve it. Engineering time, not machining time |
| Deep pocket, slender tool | Deflection rises sharply with overhang; lighter engagement, many more passes |
| Small internal corner radius | Sets the minimum cutter, which sets the achievable feed |
| Thin walls, asymmetric stock removal | Residual stress; the part moves after machining and wants stress relief sequenced in |
The quantity, which reframes everything above it. Programming, fixture build, tool presetting and proving the first piece are near enough identical work at five pieces or five hundred. At five, that entire burden lands on five parts and the buyer is mostly purchasing an afternoon of engineering with some components attached. At fifty it dilutes tenfold. At five hundred it disappears into the noise and the price has flattened onto the true cost of cutting, below which no order size will take it. Most pricing arguments in job shop work trace back to this one misunderstanding, and the remedy costs nothing: state the annual offtake even when ordering in tranches, because a single setup can serve scheduled releases if that's known in advance.
The notes column, routinely the most expensive line on the sheet. Full dimensional inspection with recorded results is a different exercise entirely from sampling. A first article package (bubbled drawing, actual measured values against every characteristic, material and process certifications attached) is genuine hours, amortised across the batch exactly like setup, so short runs feel it hardest. CMM programming for a new part is its own setup cost. And a single line sending the job out for heat treatment, plating or NDT brings batch minimums and dwell times that on low volumes can rival the machining itself.
Five minutes with one sheet of paper, and the number is broadly decided. Which is the useful part: cost in precision machining isn't argued down afterwards, it's designed in beforehand, and a drawing reviewed before it's frozen is worth more than any quotation reviewed after it lands.
Which is exactly what design for machining is for. The awkward economics of any component are that the decisions committing most of its cost are made in the first few weeks, while almost none of the money has been spent, and by the time a drawing is released, revision-controlled and quoted, the cheap window for changing it has closed. Moving a corner radius on screen takes an afternoon. Moving it after release means a revision, a re-quote, possibly a fresh first article, and if it's already in production, an operation carried on every part for the life of the programme.
DFM isn't about compromising a design to suit a machine. It's about not paying for things the function never required. Most of the savings come from a small, unglamorous set of moves:
- Tolerance only what does work. A handful of dimensions mate, seal, locate or carry load. The rest exist so the part has a shape, and tolerances inherited from a template five revisions ago cost real money.
- Match radii to standard cutters. An internal corner sized to a cutter already in the carousel is free. One sized 0.5 mm smaller demands a special tool, a slower feed, or both.
- Keep datums local where function allows. Referencing everything to one distant datum is tidy on paper and forces either a single setup or an expensive fixture.
- Watch depth against tool diameter. Deep, narrow features are a rigidity problem long before they're a machining problem.
- Specify the relationship, not three tight features. If what matters is how two bores relate to each other, say so. Tolerancing each of them tightly buys the same result at three times the price.
The pattern is that none of these need the design to get worse. They need it to be reviewed by someone who knows what each callout triggers downstream, at a point where changing it is still free.
That's the whole argument for sending a drawing out as a design review rather than an RFQ. A quotation tells you what a part costs. A design review tells you what it needs to cost, and the gap between those two numbers is usually larger than anything negotiation will ever recover.