The terms came out of large consumer electronics companies, where each gate can involve thousands of units. A startup doesn't need that scale. It does need the discipline: each gate asks a different question, and skipping a question doesn't make it go away. It makes it show up later, at a higher price.
Before EVT: the prototype stages
Before formal validation, you have prototypes that answer narrow questions.
- Proof of concept — dev boards and breadboards. Does the core function work at all?
- Works-like prototype — custom or semi-custom electronics that behave like the product but look like a lab experiment.
- Looks-like prototype — a model that looks and feels like the product with no working electronics.
EVT begins when works-like and looks-like merge into one integrated unit built to the intended design.
The three gates at a glance
|
EVT |
DVT |
PVT |
| Question |
Does the design work? |
Does it survive the real world? |
Can the factory build it? |
| Built by |
Engineering team, prototype vendors |
Production-intent processes |
The actual production line |
| Startup unit count |
5–20 |
20–100 |
100–500 |
| Enclosure |
Printed or machined |
Production tooling (first shots) |
Production tooling, tuned |
| Typical focus |
Functions, power, firmware stability |
Environmental, drop, thermal, certification |
Yield, cycle time, test coverage |
| Exit when |
All functions pass across all units |
All reliability and compliance tests pass |
Yield is stable at target |
Large companies use bigger numbers at every stage. The counts above fit a startup taking its first product to market.
EVT — Engineering Validation Test
EVT units should be electrically and functionally complete. The goal is to find every design bug while change is still cheap.
What gets tested:
- Every function against the requirements document.
- Power consumption in each operating state, and battery life where relevant.
- Thermal behaviour under maximum load.
- Radio performance — range, connection stability, interference.
- Firmware stability over long runs, not just a demo loop.
- Unit-to-unit variation. Five units that all behave the same matter more than one perfect unit.
Exit criteria: every requirement verified, all critical bugs fixed, design changes for DVT documented. Expect at least one more PCB revision after EVT. That's normal.
DVT — Design Validation Test
DVT moves to production-intent parts. Enclosures come from the first shots of production tooling, not printers. Boards use the final stack-up and final components.
What gets tested:
- Environmental: temperature cycling, humidity, and exposure relevant to where the product lives.
- Mechanical: drop tests, button and connector cycle life, ingress protection if claimed.
- Compliance: pre-compliance and formal testing for radio emissions, safety and electrical standards in each market.
- Field trials: real users in real environments.
Field trials catch what labs miss. Our automated hydroponics system runs in a live hospitality environment. Real water chemistry, real humidity and real staff behaviour expose issues no bench test reproduces. DVT is where you invite that reality in on purpose.
Exit criteria: all reliability and compliance tests passed, design frozen. Changes after DVT should be rare and painful by design.
PVT — Production Validation Test
PVT is the factory's exam. The design is frozen. The question now is whether the production line can build it repeatably.
What gets measured:
- First-pass yield — the share of units that pass all tests first time.
- Cycle time — how long each unit takes to build and test.
- Test coverage — whether the line test catches every defect type you've seen.
- Packaging and shipping — whether units survive transport.
PVT units are often sellable. Many startups use them as the first customer shipment.
Exit criteria: yield stable at target, defects understood with root causes, factory documentation complete.
The mistakes we see most
Entering EVT with a looks-like model glued onto a works-like board. If the integrated unit doesn't exist yet, you are not in EVT.
Changing the design during DVT. Every change resets some of the validation work. Batch changes and accept that some ideas wait for version two.
Skipping pre-compliance. A formal radio test failure in DVT can force a board respin and new tooling. A pre-compliance scan during EVT costs a fraction of that.
Treating PVT as a formality. The factory will find problems your engineering team never saw. Budget time for them.
How long each gate takes
For a startup product of moderate complexity, EVT runs four to eight weeks, DVT six to twelve weeks — tooling lead times dominate — and PVT three to six weeks. Certification can run in parallel with late DVT.