NPI: from working design to repeatable process

What NPI is, and what it is not

New Product Introduction is the work between a design that functions on the bench and a manufacturing process that produces it repeatably, at the quality and cost the product needs. A prototype proves that the circuit works. NPI proves that the factory can build it — the same way, run after run, with a documented way to know when something has drifted.

NPI is not redesign. It produces findings and options for the design owner; decisions stay with you. It is not a single build; it is a sequence of reviews and controlled builds with exit criteria. And it is not optional for anything that will ship in volume. Teams that skip it do not avoid the work; they do it later, on production boards, at higher cost.

Where NPI sits: EVT, DVT, PVT

Hardware teams commonly describe validation in three stages. NPI activities run through all three, shifting emphasis from design to process as the product matures.

The labels vary between companies. What matters is the sequence: prove the design, then prove the process, and do not lock the second before the first is stable.

Design-for-manufacturing review

DFM review concerns the bare board: can the fabricator build it reliably, at good yield, to the specification? The review is performed against the selected facility’s actual capabilities, which is why the facility should be chosen before the review rather than after. Typical findings include:

• Trace width or spacing below the facility’s standard capability, forcing a premium process or a design change.
• Insufficient annular ring, where drill tolerance can break out of the pad.
• Drill-to-copper clearances that risk shorts after lamination shift.
• Solder mask slivers too thin to survive processing, and mask openings that expose adjacent copper.
• Acid traps and sharp acute angles in copper.
• Copper distribution that will warp the panel, or copper too close to the board edge for routing.
• Impedance-controlled traces whose target cannot be met with the specified stackup.
• Via-in-pad without fill and cap, which wicks solder away from the joint.

Findings are dispositioned one of three ways: change the design, accept the risk with a documented rationale, or change the process (a different facility, class, or material). The disposition is recorded so the next revision does not reopen it.

Design-for-assembly review

DFA review concerns placing and soldering components. It uses the BOM, the pick-and-place file, the assembly drawing, and the paste layer. Typical findings:

• Footprints that do not match the component’s recommended land pattern — the most common cause of tombstoning and open joints.
• Components too close together for placement or rework tooling, or too close to the board edge for the conveyor.
• Missing or poorly placed fiducials, so the machine cannot align the board.
• Paste apertures that are wrong for the pad size, especially on fine-pitch and bottom-terminated packages.
• Tall components shadowing smaller ones in reflow, or through-hole parts that block wave soldering of neighbors.
• Polarity ambiguity on the silkscreen and drawing.
• Mixed-technology sequences that require hand soldering when a small change would allow a machine process.
• Test-point coverage insufficient for the intended test method.

DFA findings are usually cheap to fix in layout and expensive to fix on the line. The review should happen while the layout is still open.

Revision lock and change discipline

At some point the design stops moving and the process starts to be built around it. That point is the revision lock. It is a documented decision: this revision of fabrication data, this BOM, this assembly drawing, this test procedure. Everything downstream — stencils, programs, fixtures, documentation — references it.

After the lock, changes go through an engineering change process rather than being absorbed. A change request states what changes, why, what it affects (fabrication, BOM, assembly, test), and when it takes effect. It is approved by the design owner and released as a new revision. Work already in progress is completed to the current revision unless the change is safety-related. This discipline feels bureaucratic on a small team and becomes indispensable the first time a field failure has to be traced to a build.

The build documentation set

Repeat production depends on a set of documents that fully describe how to build the product. NPI produces or completes them:

• Released fabrication data and fabrication drawing, with the stackup as built.
• Controlled BOM with approved alternates and no-substitute lines, matched to the pick-and-place file.
• Assembly drawing with polarity, DNP callouts, and special instructions.
• Process notes: reflow profile constraints, wash or no-wash, coating, torque values, sequence of sub-assemblies.
• Inspection plan: what is inspected, by which method, to which criteria, with photographs of acceptable and unacceptable conditions where useful.
• Test procedure: programming steps, functional tests, acceptance limits, fixture and software versions.
• Packaging and labeling specification.
• Change history and the dispositions of DFM and DFA findings.

If a new facility could build the product correctly from this set alone, the documentation is complete. That is the practical test.

First-article inspection

The first units from a new process — a new revision, a new facility, or a significant process change — are inspected against the drawings and specification before the remainder of the run proceeds. First-article inspection checks that the build matches the documentation: correct parts at correct locations and orientations, solder joints to the agreed acceptance class, board dimensions and finish per the fabrication drawing, programming and functional test passed.

The outcome is a disposition: proceed, proceed with a documented deviation, or stop and correct. First-article results are part of the build record, and they become the reference for what a correct unit looks like when a question arises later in the run.

The pilot build

A pilot build is a controlled quantity built on the production process at the production facility. Its purpose is to measure the process, not to ship product — although pilot units frequently do ship once they pass. Quantity is chosen to expose process variation: enough boards that a defect rate becomes visible, not so many that a systematic problem is expensive.

What the pilot measures: first-pass yield at each inspection and test step; defect types and their locations on the board; time per unit at each operation; component attrition; and any instruction the operators found ambiguous. Each finding either changes the documentation or is accepted with a rationale. The pilot is complete when the process produces the expected yield and the documentation reflects what was actually done.

Developing the test strategy

Prototypes are tested on the bench by the people who designed them. Production cannot be. NPI is where bench knowledge becomes a test procedure that an operator or a fixture can execute. The progression is usually: define what “working” means as a list of observable checks; decide the method for each (programming success, power-on self-test, communication check, measured parameter); determine whether a fixture is needed or whether the board’s own connectors suffice; and set acceptance limits with margin for production variation.

Test coverage is a trade-off between confidence and cost. In-circuit test finds assembly defects (opens, shorts, wrong values) but needs a fixture and test points. Functional test finds whether the product works but may not localize the fault. Many products use programming plus a functional check at low volume and add fixtures as volume justifies them. The decision belongs in the NPI plan, not in the first production order.

How NPI goes wrong

• Locking the revision before the design is stable, then absorbing changes informally.
• Running DFM against generic rules instead of the selected facility’s capabilities.
• Treating the pilot as a production order and shipping before the findings are reviewed.
• Fixing problems on the line without updating the documentation.
• Leaving test to the first production run.
• Approving substitutions verbally, so the next build uses a different part.
• Changing facilities after the pilot without repeating first-article inspection.
• No named owner on the manufacturer’s side, so findings are scattered across vendors.

Exit criteria for repeat production

The design is ready to leave NPI when all of the following are true: the revision is locked and released; DFM and DFA findings are dispositioned and closed; the build documentation set is complete and reflects the pilot as built; first-article inspection passed on the production process; pilot yield meets the target with defects understood; the test procedure is documented with acceptance limits; and the approved-alternate list covers the lines most likely to be short. From here, repeat production is a matter of ordering to the controlled revision — which is exactly the point.