Key takeaways
- NPD asks whether the product can exist. NPI asks whether it can be built repeatedly.
- Neither term is standardised, which is why two vendors can both claim both and mean different scopes.
- The seam between them is design transfer: the moment the design stops being the design team’s and becomes the factory’s.
- Programs stall at that seam because responsibility changes hands exactly where the remaining risk sits.
NPD is the work of making a design that functions. NPI is the work of making that design buildable, repeatedly, at volume, by people who did not design it. New product development ends with a device that works on a bench. New product introduction ends with a production line that turns out the same device a thousand times without you in the room. They are sequential phases, they need different skills, and they are frequently sold by different companies.
Neither acronym appears in the standards that govern staged development. ISO 9001, ISO 13485 and AIAG’s APQP all define phased design and development without using either term. That matters more than it sounds, because two vendors can use “NPI” to mean quite different scopes and both be using the term correctly.
Two phases, often two different vendors
| NPD (new product development) | NPI (new product introduction) | |
|---|---|---|
| Question it answers | Can this product exist? | Can this product be built repeatedly? |
| Main outputs | Working design, validated architecture, released CAD and schematics | Process documentation, tested tooling, yield data, a qualified line |
| Typical owner | Engineering or design firm | Contract manufacturer, or a manufacturing engineering team |
| Measure of success | It works | It works every time, at cost, at yield |
| Where DFM sits | Should start here | Usually where it actually starts |
Phased development itself is well established, even though the NPD and NPI labels are not standardized. The academic anchor is Robert G. Cooper’s “Stage-gate systems: A new tool for managing new products,” Business Horizons 33(3), 1990, which set out the stage-and-gate structure most hardware programs still use. ISO 9001:2015 clause 8.3 requires planned design and development stages with defined inputs, controls, outputs and change control, though it names no specific stages. In automotive, AIAG’s Advanced Product Quality Planning is the sector’s gated framework; its third edition was released in March 2024 and split the control plan into a separate manual.
What NPD covers
From concept to a design that works
NPD is where the product is invented and proven. The deliverables are engineering artifacts, and a program that cannot name them is not really in NPD:
- Requirements and architecture. What the product must do, at what cost, in what size, under what regulations.
- Industrial design. Form, ergonomics, materials and finish.
- Mechanical and electrical design. CAD, schematics, PCB layout, thermal and structural analysis.
- Firmware and embedded software. Enough to prove the product’s function, not necessarily enough to ship.
- Functional prototypes. Built to answer specific questions, not to look finished.
- A first BOM. Component-level, with alternates identified where supply is thin.
Design for manufacturing and design for assembly belong here even though they describe manufacturing concerns. The DFMA method came out of academic work by Geoffrey Boothroyd and Peter Dewhurst, whose central test is the theoretical minimum part count: a component earns separate existence only if it needs a different material or process, has to move relative to adjacent parts, or must be separate for assembly or service. Applying that test during NPD is cheaper than applying it during NPI, when the CAD is frozen.
What NPI covers
From a design that works to a design that can be built repeatedly
NPI converts a validated design into a manufacturing process. The engineering does not stop, it changes subject: from whether the product works to whether the line can make it work.
- Manufacturing process design. Assembly sequence, work instructions, fixtures, station layout.
- Test development. Functional test, in-circuit test, calibration steps, pass criteria and traceability.
- Tooling. Mold design, first-article inspection, tuning across shots.
- Supply chain qualification. Approved vendor lists, second sources, incoming inspection.
- Yield and cost work. Measuring first-pass yield, finding the stations that lose units, and closing the gap to target cost.
- Handoff documentation. The package the factory runs from once your engineers leave.
Parts of that package are standardized, which is useful when comparing vendors. IPC-2581, marketed as DPMX, defines a single-file format for transferring printed board assembly manufacturing description data, and IPC-D-356 covers bare board electrical test information in digital form. IPC-A-610 sets acceptability criteria for electronic assemblies and classifies them into three product classes in its section 1.3, and J-STD-001 covers soldered assembly requirements. One naming note: IPC became the Global Electronics Association in June 2025, and many resources now sit on electronics.org, though the standards keep their IPC designations.
EVT, DVT, and PVT
These three build stages are the vocabulary most NPI conversations run on, and they are industry convention rather than standardized terms. Even the expansions disagree between sources: EVT appears as both engineering validation test and engineering verification test, and DVT as design validation or design verification. The convention, as used across consumer electronics manufacturing:
- EVT. Early builds that ask whether the design works at all. Failures are expected and are the point.
- DVT. Larger builds using production-intent parts, run against the full specification, including compliance testing.
- PVT. A build on the real line with real tooling and real operators, to prove the process rather than the product.
The terms are convention rather than standard, so make each build’s exit criteria explicit in the contract. “PVT complete” means whatever the two of you wrote down, and nothing more.
Where the seam between them sits
The seam is design transfer: the moment the design stops being the design team’s and becomes the factory’s. For medical devices the seam is regulated. FDA’s Quality Management System Regulation took effect on February 2, 2026, amending 21 CFR Part 820 to incorporate ISO 13485:2016 by reference, so the former design controls section at 21 CFR 820.30 is now reserved and the design and development transfer requirement sits in ISO 13485 clause 7.3.8. That scope is medical devices only. No equivalent federal requirement applies to consumer hardware, which means for most products the seam is defined by your contract or not at all.
What a real transfer package contains:
- Released drawings and CAD, with revision control and a change history.
- A production BOM carrying approved manufacturers and part numbers rather than generic descriptions.
- Assembly and test instructions written for someone who has never seen the product.
- Acceptance criteria, including which IPC class applies and why.
- Test fixture designs, calibration procedures and golden units.
- Known issues, open deviations and the workarounds currently in use.
Why the seam is where programs fail
Both sides can perform well and the program can still stall at the handoff, because responsibility changes hands at exactly the point where the remaining problems are discovered. Typical failure patterns:
- The design firm’s job ended at “it works.” The contract said working prototype, so a working prototype arrived, and nobody was paid to make it manufacturable.
- The factory quoted from an incomplete package. The price assumed a BOM that turned out to have unsourceable parts.
- DFM feedback arrived after tooling was cut. The cheapest possible change became the most expensive.
- Nobody owned test. The design team assumed the factory would write the test, and the factory assumed test came with the design.
- The BOM was never customs-ready. Bills of materials carry legal weight beyond engineering; 19 CFR § 190.29, for example, requires drawback claimants to certify possession of the applicable bill of materials.
A word on a figure you will see quoted. The claim that design determines about 70 percent of product cost has been disputed in the peer-reviewed literature, in Barton, Love and Taylor’s “Design determines 70% of cost? A review of implications for design evaluation,” Journal of Engineering Design 12(1), 2001. The direction of travel is uncontroversial: changes get more expensive once tooling and supply commitments exist. The specific multiplier is not evidence you should plan against.
Vendors that do one, and vendors that do both
The phase split explains the vendor landscape, because most companies are built around one phase:
- Design and engineering firms sell NPD. A product development firm hands over documentation and stops there.
- Contract manufacturers and EMS providers sell NPI and production. An EMS company builds to documentation it did not create.
- Integrated firms sell both, which removes the handoff and concentrates dependency in one supplier. The wider comparison across company types is in OEM vs ODM vs EMS.
Neither structure is automatically better. A split gives you two sets of eyes and one seam to manage. An integrated program removes the seam and removes your second opinion with it.
What to ask a vendor that claims both
“We do NPD and NPI” is a common claim and an unfalsifiable one until you make it specific:
- Which of your people do NPI, and are they the same people who did the design? Different disciplines, and the honest answer names two teams.
- At what point in the design does your manufacturing engineer first see it? Late answers predict late DFM.
- Show me a transfer package you produced. Redacted is fine. Its completeness is the answer.
- What IPC class do you build to by default, and why would that change for my product?
- Which build stages are in scope, and what are the exit criteria for each? Get EVT, DVT and PVT defined in writing.
- Who owns the test fixtures and the process documentation when we finish? This is a contract question and worth an attorney’s eye alongside the IP terms.
Frequently asked questions
Is NPI just manufacturing?
No. Manufacturing is the repeated production of an established process. NPI is the engineering work that creates that process, including test development, tooling qualification and yield improvement.
Can NPD and NPI overlap?
They should. Manufacturing input during design is the cheapest form of DFM available, and the alternative is discovering process constraints after the CAD is frozen.
Does my product need a formal NPI phase?
Any product with tooling, a supply chain and a test step has an NPI phase, whether or not anyone names or budgets it. Unnamed and unbudgeted is where the cost surprises come from.
Are EVT, DVT and PVT standardized?
No. They are industry convention rather than standardized terms, and even the acronym expansions vary between sources. Treat them as conventions and write exit criteria into the contract.
How long does NPI take?
It depends on tooling, certification and component lead times more than on engineering effort. The time it takes to develop a hardware product breakdown shows where the calendar actually goes.
Who owns the process documentation?
Whoever the contract says. It is separable from design ownership, and a program can leave you owning the design while the factory owns the process that makes it. Settle both in writing.
Where Inventornest fits
Inventornest handles product design and development in-house, and we coordinate directly with our manufacturing partners and provide the production package appropriate to the product and agreed scope. Our OEM services cover concept through production handoff. If you are trying to work out which phase your product is actually in, you can book a consultation and we will tell you what is missing before you commit to anything.
