OEM projects rarely fail at the drawing stage. They fail at the handovers between stages, when one side assumes the other has already checked something. A bracket dimension, a certification scope, a firmware version, a tooling ownership clause — each is small on its own, and each becomes expensive once tooling exists and a launch date is published.
The seven gates below are the points where a project either transfers cleanly to the next stage or quietly carries a defect forward. Each gate has a document that should exist before the next one starts, and a failure pattern that shows up later if it does not.
OEM and ODM are not the same project
The two words get used interchangeably in early conversations and diverge sharply afterwards. In one, the buyer brings the design and the factory builds it. In the other, the factory brings a platform and the buyer's requirements are applied to it. The difference decides who owns the drawing, who holds the certification, who pays for tooling and who is answerable when a unit fails in the field.
Settle that before a specification is written, in a table with one row per item: design, tooling, firmware, certification, packaging, field failure. Each row needs a single named owner, not a shared one. Shared ownership at this stage becomes shared blame at the end of the project, and blame does not ship units.
Gate one: requirement and responsibility split
The deliverable is a requirement document that states what the product does, what it does not do, and which market it is for. Vague entries here are the most expensive kind: "European version" can mean a connector, a language set, a metering rule, a payment method and a documentation language, and each of those sits with a different team.
The usual failure is a requirement written by sales and read by engineering. Sales describes an outcome the customer wants; engineering builds the outcome that was written down. Put both in the same review, and record the open points rather than resolving them verbally — an open point with an owner is manageable, a closed point that nobody wrote down is a change order later.
Gate two: feasibility before the first drawing
Before industrial design is frozen, three things are worth answering in writing: whether the enclosure has room for the thermal path, whether the electrical clearances survive the enclosure that marketing prefers, and whether the chosen connector and cable bend radius fit the parking layout the buyer has in mind.
The typical failure is aesthetic decisions made before these answers exist. A slimmer cabinet looks better in a render and leaves no space for airflow, so the thermal solution becomes a derating strategy, and a derating strategy becomes a conversation with the buyer's operations team about why the site is slower than the brochure. Feasibility review is cheap; the redesign after tooling is not.
Gate three: a prototype is not a product
The EVT stage produces units that prove the concept. They are often built by engineers, with hand-modified parts, outside the standard process. That is normal. The mistake is treating their test results as evidence about series production.
What the gate needs is not a passing test but a problem list with an owner and a disposition for each item: fixed, accepted with justification, or deferred to a named later gate. Deferred items are where projects lose control, because a deferred thermal or EMC issue reappears exactly when the enclosure is already tooled. Ask for the list, not the summary slide. If there is no list, there was no review.
Gate four: validation, certification and the design freeze
DVT is where the design is tested against the requirement rather than against itself. The gate has two outputs that matter commercially: a validation report that covers the operating envelope, and a certification scope that names exactly which variant was assessed.
The failure that surfaces months later is scope drift. A unit is certified as submitted, then a connector, a meter or a card reader changes for a customer request, and the certificate no longer describes what is being shipped. Freeze the configuration at this gate, and make every later change pass through the change process in gate seven. A certificate without a variant list is a document about a product that may no longer exist.
Gate five: the pilot run has to be built by production
PVT is the first build that counts, because it uses the real line, the real fixtures and the real operators. If the pilot is staffed by the development team, it proves that the development team can build the product, which is not in doubt.
What the gate should produce: a build report with the yield at each station, a list of fixtures and test equipment that exist and are calibrated, work instructions that an operator can follow without asking an engineer, and a record of every deviation made during the build. Deviations are the interesting part. A pilot with no deviations usually means nobody was watching closely enough.
Gate six: first-article approval and the control plan
Approving the first article is the buyer's last cheap opportunity to stop a problem. The article should be taken from the production line, not selected by the quality team, and it should be measured against the drawing rather than compared visually with a sample.
Alongside it, three documents set the standard for everything that follows: a control plan naming what is checked at which station, an incoming inspection standard for critical components, and a packaging specification that has survived a drop test. Without the control plan, quality depends on who is on shift. With it, quality has a definition that a new operator can inherit.
Gate seven: ramp-up and change control
Ramp-up is where volume exposes everything the pilot absorbed. A line that built a handful of units with engineers nearby behaves differently when the schedule doubles and the experienced operators are spread across two shifts. The gate needs a ramp plan with a stated review point, and a rule for what happens when yield drops below the pilot figure.
Change control is the other half. Once units are in the field, every change needs a record that maps to a serial range, a customer notification path and a decision about whether earlier units are affected. This is also the point to settle spare parts: how long modules continue to be available after production ends, and at what level — board, module or assembly. A product that cannot be repaired in its fifth year is a product with a shorter life than its warranty.
Seven gates on one page
| Gate | Document to hold | Where it usually breaks |
|---|---|---|
| Requirement | Requirement document and responsibility matrix | Two meanings of OEM; owners left shared |
| Feasibility | Feasibility note covering thermal, clearance, connector space | Industrial design frozen before engineering review |
| Prototype | Test report plus issue list with dispositions | Hand-built units read as production evidence |
| Validation | Validation report, certification scope with variant list | Configuration keeps moving after assessment |
| Pilot | Build report, station yield, deviation log | Pilot staffed by the development team |
| First article | Measured first-article report, control plan, packaging spec | Article selected rather than sampled |
| Ramp | Ramp plan, serial-to-revision map, spare-parts commitment | Changes made without a record after launch |
Five clauses belong in the contract rather than in the project plan, because they outlive the people who negotiated them: who owns the design and the tooling, who holds the certificate, how changes are notified, how long spare parts continue, and what happens to the buyer's tooling if the relationship ends. Each is a single sentence to write at the start and a dispute to settle at the end.
Projects that reach volume without drama tend to look the same: every gate produced a document that someone outside the project could read, and every open point had a name next to it. The engineering behind an AC wallbox and a DC station differs, and so does an ultra-fast cabinet, but the gates do not — and the technical documentation produced along the way is what a buyer inherits when the project team disperses.
Map the seven gates against your next project
If a gate has no document behind it, that is the item to raise at the next review — before tooling, not after launch.
Contact our team