Why Production Is Where the Expensive Errors Appear

A prototype is one unit, made by a careful process, assembled by a person who wants it to work. Production is thousands of units, made by a process with statistical variation, assembled by an operator with a 30-second cycle time who has never seen your CAD. Everything that depended on care rather than on design shows up in that gap.

The cost curve is brutal. A change made in CAD costs an hour. The same change after the mold is cut costs $2,000 to $15,000 and two to four weeks. After the first production run, it also costs the inventory. This is why the mistakes below are worth memorizing: none of them are exotic, and all of them pass a prototype review.

Mistakes in Molded Plastic Parts

Non-uniform wall thickness. A printed prototype does not care that one region is 4 mm thick and its neighbor is 1.8 mm. A molded part does: the thick section cools last, shrinks more, and pulls a visible sink mark or a warped face. The rule and its consequences are in wall thickness for injection molded parts. Core out thick sections; do not solve them with process settings.

Missing or insufficient draft. Zero-draft walls print perfectly and will not eject from steel without drag marks or without an ejector punching through. One to two degrees on most surfaces, three or more on textured ones, per draft angles and parting lines. Adding draft late changes every mating dimension, which is why it cascades.

Ribs that are too thick. A rib at the full wall thickness looks solid in CAD and produces a sink mark on the show surface directly opposite. Keep ribs at 50 to 60 percent of the adjoining wall.

Knit lines through structure. Where two flow fronts meet behind a hole or a boss, the material fuses at 50 to 80 percent of full strength. A prototype machined from billet has no such line. If a snap arm or a mounting boss sits downstream of an opening, it may snap in production and never in prototyping.

Undercuts nobody costed. A small lip that a printer produces for free requires a side action or a lifter, adding $3,000 to $12,000 to the tool and often a parting line across a cosmetic face.

Tolerance Stack: The Assembly That Will Not Close

The most common production surprise is a set of parts that are all individually within specification and still will not go together. Each part was checked alone; the chain was never checked.

Prototypes hide this because you make one of each, and one of each nearly always fits. Production makes a thousand of each, and eventually a part near the top of its tolerance band meets one near the bottom. If eight features stack in a line and each carries plus or minus 0.15 mm, the worst case is 1.2 mm of accumulated error, which is enough to prevent a lid from closing. Running the analysis in tolerance stack-up analysis before release costs a day.

Two related errors travel with it. The first is toleranced dimensions with no stated datum, which lets the factory measure from wherever it likes and still pass; the fix is proper drawing practice as covered in what a manufacturing drawing must include. The second is putting tight tolerances everywhere as insurance, which inflates the quote and teaches the factory to ignore all of them.

Assembly Errors That Only Appear at Line Speed

An engineer assembling a prototype has all afternoon and full knowledge of the design. An operator has 30 seconds and a work instruction. Designs that require patience fail at that speed.

  • Parts that assemble two ways. If a gasket, bracket, or PCB can go in backwards, some percentage will. Add an asymmetric feature so only one orientation physically fits.
  • Fasteners you cannot reach. A screw that needs a 90-degree driver or a hand inside the housing turns a 5-second operation into a 40-second one and multiplies labor cost across the run.
  • Wires with no strain relief or routing. Loose harnesses get pinched under covers. Some will be cut through and fail in the field. Design a channel and a clip.
  • Too many unique fasteners. Six different screw types means six bins and six chances to grab the wrong one. Standardize, and reduce part count using the principles in design for assembly.
  • No way to test before final close. If the product is glued or ultrasonically welded shut before functional test, every failure becomes scrap instead of rework.
  • Assemblies that need a third hand. Holding two parts aligned while starting a screw is fine once and terrible ten thousand times. Add a locating feature that holds the parts until the fastener lands.

The Quiet Category: Field Failures Designed In

Some mistakes clear production and arrive later as warranty claims. Plastic threads stripped by an operator's power driver. Snap features that survive assembly and crack after twenty user openings. Dissimilar metals that corrode in a humid climate. Parts that pass at room temperature and bind at 100 F. All share one root cause: the design was validated once, gently, by someone who knew where the fragile parts were.

How to Catch Them Before the Tool Opens

Four checkpoints catch the great majority.

A real DFM review with the actual manufacturer, not a generic checklist. Send the CAD to the molder or machinist and ask what they would change; they do this daily and their notes are free. The framework is in design for manufacturing.

A mold flow simulation on any part above roughly $8,000 of tooling. It costs $500 to $2,500 and predicts warp, sink, knit line location, and fill problems while steel is still uncut.

A written tolerance analysis on every assembly chain that must close, with a stated decision on whether you are designing to worst case or to a statistical distribution.

An assembly dry run. Print or machine a full set, hand them to someone who did not design the product, give them the work instruction and nothing else, and time it. Every hesitation is a design defect.

Then treat the first shots as data rather than as product, using the inspection sequence in T1 samples, and run a proper pilot production run before committing to full volume. The pilot is where line-speed problems become visible, and it is far cheaper to learn there than in a container of finished goods.

Have the Design Reviewed Before the Steel Is Cut

Projects House runs pre-tooling design reviews that cover moldability, tolerance chains, and assembly sequence, and returns a prioritized change list your factory can act on in one revision. Send your CAD and the manufacturing process you plan to use through our contact form.