Where OEM Product Development Goes Wrong

The pattern behind most delayed projects

A project rarely fails where it appears to fail.

The EMC pre-scan that comes back red two weeks after the first engineering build looks like a testing problem. It is almost always a design decision made months earlier. The tooling that needs a third revision looks like a mold problem. It is almost always a specification that was never frozen. The launch that slips a quarter waiting on one component looks like a supplier problem. It is almost always a bill of materials that nobody graded for risk.

This matters because of where the money is. Changing a drawing costs an afternoon. Changing steel costs weeks and five figures. Changing a product that is already in customers’ hands costs a recall. The cost of a fix roughly follows how far downstream you catch it, so the useful question is not “how do we move faster” but “where is this project accumulating risk that nobody has looked at yet.”

What follows is the five-stage sequence we run, and the failure that shows up in each one.

Stage 1. Positioning: the specification that never froze

Most projects open with a product idea and a target price. What they often lack is the two numbers that determine almost everything downstream: expected annual volume and target market.

Volume decides cavity count, and cavity count decides tooling cost. Quote a single-cavity tool against million-unit expectations and the piece price never works; quote eight cavities against twenty thousand units and you have paid for capacity that never runs. Target market decides the certification path, and the certification path decides which components are even eligible.

When those two answers arrive late, the specification keeps moving, and every stage after this one inherits the movement. The most expensive projects we see are not the technically hard ones. They are the ones where nobody would commit to a number in the first month.

→ For how cavity count, mold steel and changeover economics turn into a quote: Tooling Cost and MOQ Explained for OEM Projects

Stage 2. Engineering review: the risks nobody scored

This is the stage where a project either surfaces its problems or buries them.

We run a structured design-for-X review before any steel is cut: 33 check items, each rated red, amber or green, and a single red stops the project. Not “flag for later.” Stops.

On one portable UV-C soft-bag product, that review returned six reds. Among them: EMC pre-scan needed to move to two weeks after the engineering build rather than before mass production; the waterproof zipper needed a 10,000-cycle open-close test; IEC 62471 light-leakage testing had to be scheduled; a reed switch plus firmware double interlock was required rather than one safety path; a 275nm UV-C LED on an 8 to 12 week lead time needed a second source; and a direct USB supply had to be reworked because it sidestepped UN38.3 and IEC 62133 entirely.

Every one of those would have surfaced anyway. The difference is whether it surfaces on a review sheet or in a certification lab after the tool exists.

→ The full method and case: DFX Risk Review Before Cutting Tooling

Stage 3. Tooling and cost: the component that holds everything

By this stage the design is committed. The failure mode changes shape: it is no longer about what the product is, but about whether every part of it can actually be bought, repeatedly, on schedule.

We grade the bill of materials before tooling. On that same UV-C project the grading came out as: the 275nm UV-C LED red: special specification, 8 to 12 week lead time, single source; the UV-resistant reflective fabric and the fan amber; the waterproof zipper green, single brand but stable, ordered six weeks ahead; the PCBA and MCU green, two suppliers across two platforms.

One red component sets the schedule for the entire product. Nothing else about the build matters if that part arrives in week twelve.

→ The four grading axes and what each level triggers: Component Supply Risk Matrix for OEM Projects

Stage 4. Packaging and pilot: the first run that is not a sample

Pilot production is where a design meets people who did not design it. Assembly sequence, fixture design, operator instructions and packaging all get exercised for the first time at something resembling real volume.

The failure here is subtle: everything technically works, but the build is not yet repeatable. A step that a senior engineer performs correctly by instinct becomes a defect when it moves to a production line. This is the stage that decides whether your yield at unit 10,000 looks like your yield at unit 100.

Stage 5. Quality assurance: the loop that has to close

Inspection at incoming, in-process and final shipment, plus third-party testing where the market requires it. But the stage does not end at shipment. It ends at whether the factory can act on what comes back from the field.

One home-use electronic device generated a complaint about scorch marking at the base. The investigation traced it to USB Type-C overheating, which produced an engineering change from a 26-pin to a 6-pin configuration plus current detection, then a horizontal review across related products, then batch traceability to identify what had already shipped.

That sequence (complaint, root cause, engineering change, horizontal deployment, traceability) is what an ISO 13485 corrective and preventive action loop is for. A factory without it will fix your unit. A factory with it will find the other units with the same problem.

→ The full loop: Field Complaint to Design Change: A CAPA Loop

What to ask a prospective manufacturer

Three questions worth more than a capability brochure:

1. “At which stage gate would you stop my project?” A partner who has never stopped a project has never protected anyone from one.
2. “Which component on my BOM worries you most, and why?” This tests whether anyone has actually read your design.
3. “What happened the last time a product came back from the field?” The answer reveals whether corrective action is a process or a promise.

We build under ISO 13485:2016 and ISO 9001:2015 quality systems, with engineering, tooling, PCB/PCBA and assembly on one site in Taichung, Taiwan. Engineering changes do not queue behind a subcontractor, and when a ramp-up gets difficult there is nobody else to point at.

NDA first, and an engineer, not a sales script, replies within two business days.