Prototype to production: the six-week gap

A prototype CNC machined bracket costs 180 euros and ships in four days. The 500-unit production run is quoted at 11 euros per piece and eight weeks. Three months later the parts are still not shipping. The shop is not lying, the operator is not slow, and the design is not wrong. The gap between those two numbers is almost always the same gap, and it has a specific shape.

The prototype was machined by the shop's senior technician on a 3-axis mill from a single bar of 6061-T6. He read the drawing generously. Where it said "±0.2mm", he held 0.05mm because a manual machinist working on one part at a time naturally holds tighter than the tolerance called out. The finish was hand-sanded before anodizing. The anodize job went into a fresh tank on a Tuesday morning. The result looked and measured perfect.

Production runs differently in every one of those dimensions.

What changes between one and five hundred

Five hundred parts go through a production cell, not a prototype cell. The drawing is read literally, not generously. The stock comes from multiple lots across multiple weeks, each within spec on the mill certificate but varying slightly in grain structure and hardness. The anodize tank is six weeks old, with accumulated drag-out and slightly different chemistry than the tank the prototype went through. Two operators rotate through the work cell across three shifts. A fixture holds the part while the spindle runs at production feed rates, not prototype feed rates.

Each of those changes is individually fine. Together they explain the six-week gap.

The three gaps that eat the weeks

1. Tolerance stack-ups that parse differently at scale

A drawing with "±0.2mm" across every feature works for prototyping because the technician holds tighter on the features that matter. In production, the operator holds to what is printed. If three features each have ±0.2mm tolerance and they stack into an assembly, the worst-case misalignment is 0.6mm. At prototype scale that misalignment never appears, because the senior machinist quietly corrects it. At 500 units with rotating operators, 3 to 5 percent of parts will fail assembly on the first fit-up, and the shop will not know why.

The fix is GD&T callouts with reasons. Not every feature needs a tight tolerance. The ones that carry function (mating features, datum references, bearing seats) need ±0.05mm with positional tolerance. The ones that do not (cosmetic edges, non-mating faces) can open to ±0.25mm or looser. Every feature gets a tolerance. Every tolerance gets a reason. The ones without a defensible reason get opened, which saves money.

This audit costs between 200 and 600 euros of engineering time if done before tooling is cut. It costs 15,000 to 40,000 euros in tooling rework if discovered after first article inspection.

2. Finishes that pass on five parts but drift on fifty

Anodizing is the typical offender. Type II anodize at a target of 15 to 25 microns thickness, dyed to a specific color, looks consistent on a sample of five parts because they all went through the same bath at the same time. Across a 500-unit run, the bath chemistry drifts, the dye concentration shifts, and the rack loading varies. Without a written spec defining acceptable color delta (typically ΔE ≤ 2.0 against a master sample), the shop has no standard to hold to. Color rejects then cost between 8 and 24 euros per part in rework.

The same pattern applies to powder coat thickness (60 to 120 micron target with a 20 percent tolerance band), bead blast texture (consistency across rack position), and passivation (time in bath, acid concentration). The prototype does not surface these because the sample size is too small.

The fix is a finishing spec attached to the drawing: coating type, thickness range, color standard (Pantone, RAL, or a physical master plate), and acceptance criteria. Costs nothing to write. Saves weeks of "looks fine" versus "not fine enough" back-and-forth.

3. Assembly sequences that assumed the prototype path

The prototype went together by hand, on a bench, with the engineer watching. Fasteners threaded in whatever order felt natural. The production cell assembles in fixtures, in a defined sequence, at 60 to 90 seconds per unit. A fastener that threaded easily last when done by hand might cross-thread first when done by an operator following a sequence sheet, because the earlier fasteners have pulled the assembly slightly out of true.

The fix is a pre-production assembly audit. Run 10 to 20 units through the production cell, using the production fixtures and sequence, before tooling is finalized. Track where each operator gets stuck. Parts that fight the assembly get flagged and revised. This step adds 3 to 4 days upfront and removes 3 to 4 weeks of emergency tooling revision after first article.

What it looks like to front-load this

The three checks above, done in the 2 to 3 weeks before tooling is cut, cost roughly the same as one day of production stoppage in week 8.

1. Tolerance audit. Every dimension gets a reason. GD&T stamped on the critical-to-function features. Non-functional features opened to save cost. Deliverable: a revised drawing package.

2. Finish spec written and attached to the drawing. Coating type, thickness, color standard, acceptance criteria, sample retention requirement. Deliverable: a one-page finish addendum.

3. Pre-production batch of 10 to 20 units. Same fixtures, same sequence, same operators. Inspected against a written plan with AQL 1.0 for critical features and 2.5 for major ones. Deliverable: first article inspection report and a tooling adjustment list.

Across the programs we've audited, the average gap between first production quote and shipped first delivery runs 10 to 14 weeks when these steps are skipped, versus 6 to 8 weeks when they are done. The cost delta is typically 30 to 50 percent, driven mostly by scrap and tooling rework.

The point

The six-week gap is not a factory problem. It is a specification problem. The operators who ship on the original timeline are the ones who treated the prototype as a conversation with the future factory, not a finished design. The prototype passes because one person holds it loosely. Production passes only if the drawing holds the process tightly.

If you are about to quote a production run and the drawing you sent for the prototype is the same drawing you are about to send the factory, stop. Budget two weeks. Do the three audits above. Ship on time.

This is the kind of upstream work Sendspec does for founders before a single unit gets tooled. If you have a production run quoted and want a tolerance audit and finish spec review before you cut tooling, request a quote. We turn around a full drawing package review in 48 hours.

See also: CNC machining, die casting, aluminum fabrication.