From prototype to production: EVT, DVT, and PVT explained

EVT, DVT, and PVT are the names for the builds between a working prototype and a product coming off a production line. The terms are most associated with large consumer electronics companies and the contract manufacturers that build for them, so you will hear them as soon as you start talking to a factory.

The labels matter less than the idea behind them. Each build answers one question, and it is much cheaper to answer it in its own stage than in the next one.

The stages at a glance

1. Prototype

   Can it work at all?

   Exit when

   The architecture and the risky parts are proven, and the requirements are written down.

2. EVT

   Does the design meet its requirements?

   Exit when

   All functions work on real boards, the major bugs are fixed or understood, and the design is ready for production tooling.

3. DVT

   Does the finished product hold up?

   Exit when

   Reliability and pre-compliance pass on parts from production tools, and certification testing is under way or done.

4. PVT

   Can the factory build it repeatedly?

   Exit when

   The line runs at rate with production fixtures and processes, and the yield and quality targets are met.

5. Mass production

   Does quality hold over time?

   Ongoing

   Yield tracking, failure analysis, and controlled changes.

Before EVT: prototypes and proof of concept

Early prototypes answer “can this work at all?” They are often development kits, breakout boards, 3D-printed housings, and hand-wired circuits. That is fine. Their job is to retire the biggest risks quickly: a sensor that may not be sensitive enough, or a battery that may not last.

Before leaving this stage you should have:

• A written description of what the product must do, and where it will be sold.
• An architecture you have tested where it was most uncertain.
• The main parts chosen, with availability checked.

Early prototypes are also where “works-like” and “looks-like” models split. A works-like build proves the electronics; a looks-like build proves the size, shape, and feel. Neither is the product yet.

EVT: engineering validation

EVT is the first build of the real design: custom boards, intended parts, and a close approximation of the enclosure. The question is whether the design meets its requirements on real hardware.

What goes in. Boards built to the intended design, often in small quantities and sometimes hand-assembled. Enclosures from 3D printing, machining, or soft tooling (quick, low-cost molds that make parts close to production but not identical). Before your first board spin covers how to get the most out of these boards.

What gets tested. Every function, every interface, and the power budget. Firmware bring-up happens here. Early compliance checks start too: near-field scans and ESD on the bench, so emissions and immunity problems surface while the layout is still easy to change. See Plan for compliance from day one.

Typical exit criteria. All functions demonstrated, the major bugs found and fixed or understood, and the mechanical design stable enough to commit to production tooling. It is common to run more than one EVT build before that is true.

DVT: design validation

DVT asks whether the finished product holds up. The design should be nearly final, and the parts should come from production processes: boards from the intended assembler, and plastic or metal parts from production tooling that has been through its first trial shots and adjustments, which is usually ordered once EVT shows the mechanical design is stable.

What gets tested. This is where the product is pushed hard:

• Reliability and environmental testing: temperature, humidity, drop, vibration, and wear on buttons, hinges, and connectors, chosen to match how the product is used. For outdoor products, include time in real sun: some faults only appear there.
• Pre-compliance at a test lab, then formal certification testing on units that represent production.
• Battery life, thermal behavior in the real enclosure, and radio performance in the final housing.
• Cosmetics and fit, on parts from the real tools.

Test starting up at each temperature, not only running through the range. A product that is powered on at room temperature and then taken cold has proved that it runs in the cold, not that it starts there. I once had a switching supply that would not start between -5 and 0 C. It started fine above and below that, and once running it held through that window without trouble. Marginal component values and capacitance falling in the cold usually make a board get steadily worse as it gets colder. This did not: it was a narrow band, so I suspected the controller itself, built a clean reproduction, and took it to the manufacturer, whose own testing confirmed it was a silicon issue. I designed it out and moved on. When the evidence points at a part, a minimal, repeatable test case is what gets a manufacturer’s applications team to take it seriously.

Typical exit criteria. Reliability and pre-compliance results pass, or the fixes are known and verified. The design is frozen: changes after this point go through a formal change process, because they can affect tooling, certification, and the factory’s instructions.

DVT is where a design review pays off most visibly. A problem found in a review before DVT costs an edit; the same problem found after production tooling may cost a tool change. What I look for in a hardware design review covers what a good review checks.

PVT: production validation

PVT asks whether the factory can build the product repeatedly, at the intended rate, with the intended quality. The design should not change here. What is being tested is the process.

What goes in. Production tooling, production fixtures, the final assembly instructions, trained operators, and the real packaging. The units are built on the production line, not by engineers.

What gets tested. Line yield, test coverage, and the time each step takes. PVT units are often the first ones sold, so quality checks, labeling, and packaging must be ready. First article inspection confirms that parts and assemblies match the drawings, and packaged units go through drop and transit testing.

Typical exit criteria. The line runs at rate, the yield meets the target you agreed with the contract manufacturer, and failures are understood. Many teams keep a “golden” unit or sample set from this stage as the reference for what production should look like.

Whatever the size of the team, end each build with a short review: what was tested, what failed, the root cause, and who owns each fix.

Who is involved

Your engineering team owns the design at every stage. The contract manufacturer’s role grows with each build: they often assemble the EVT boards, and from DVT on they are a full partner whose process engineers review the design for manufacturability, build the line, and run the builds. Bring them in before EVT if you can. Their feedback on part choices, panelization, and test access is cheapest to act on before the layout is settled.

Mechanical suppliers and test labs join at different points, and someone has to coordinate them all. On a small team, that is often the founder.

What the factory needs from you, from BOM to test specification, is covered in What a complete manufacturing package includes.

Test equipment and fixtures at each stage

The right equipment makes each build faster and its results more trustworthy.

• EVT: bench instruments for bring-up and characterization, plus simple fixtures such as a programming jig and a cable harness that reaches every interface. Scripts that automate repetitive checks save days.
• DVT: access to environmental chambers and drop or vibration equipment, usually through a lab, and a first version of the production test fixture. Use DVT units to prove the fixture as well as the product.
• PVT: the production test fixture itself, typically a bed-of-nails or functional tester at the contract manufacturer (flying-probe test often stands in for a bed-of-nails fixture at low volume), with its software, limits, and logging. It must catch real defects without rejecting good units.

Production test is often left late. If test points and a test mode are not designed in by EVT, the fixture is harder to build and less thorough. Designing for production test covers what to design in.

What founders commonly underestimate

• Lead time for tooling. Production molds take weeks to build and several rounds of trial shots to adjust. Plan the mechanical design so tooling can start as soon as EVT shows each part is stable, and keep late changes “steel-safe”: designed so the toolmaker removes metal from the mold rather than adding it.
• The number of units. Each stage needs units for testing, firmware, certification, reliability (some of which are destroyed), and demos. Ordering too few means a second, small, expensive run.
• Certification time. Lab schedules, retests, and paperwork can take longer than the testing itself.
• The change process. After design freeze, every change needs to be checked against tooling, certification, and documentation.
• Production test. Fixtures and their software are a project of their own.

Adapting the stages for a small team

The full sequence was built for very large volumes and large teams. A small team rarely needs every build in its full form. Common adaptations:

• Merge EVT and DVT when the mechanical design is simple or off the shelf, such as a standard enclosure with no custom tooling.
• Use a pilot run as PVT for low volumes: a small first production build at the contract manufacturer, with the full test process, before committing to a larger order.
• Run builds in parallel where the risks are independent, for example mechanical tooling on one track while firmware matures on the EVT boards.
• Use bridge tooling (often aluminum soft tooling run in the production material) to make a first batch of enclosure parts before committing to full production steel, when volumes are uncertain.

What should not be skipped is the question each stage answers. If you merge stages, make sure the merged build still proves the design works, holds up, and can be built repeatedly. The cost of skipping a question is the same whether or not you used the label.

If you are planning the path from prototype to production and want help setting up the builds, get in touch.