EVT proves the custom electronics work. DVT proves the final design survives real use and passes certification. PVT proves the factory can build it repeatably at yield. At each gate you should receive working units, test reports and the source files behind them, and you should be able to stop and leave with everything you paid for.
Software teams know how to ship in sprints. Hardware forces a different rhythm, because every design change after a certain point costs tooling, test time and weeks. The industry splits that risk into three validation builds: EVT, DVT and PVT. This guide explains what each build is for, what a software company should receive, and what leaving looks like at each point.
What do EVT, DVT and PVT actually mean?
Each build answers one question. If you cannot say yes to that question, you should not move to the next build.
| Build | Question it answers | What is built | What may still change |
|---|---|---|---|
| EVT Engineering validation test | Does the custom electronics design work? | First custom PCB, often in a 3D-printed or CNC housing. Firmware brings up every peripheral. | Schematic and layout fixes. A second EVT spin is common. |
| DVT Design validation test | Does the final design survive real use and pass certification? | Production-intent PCB in parts from the first molds. Units go through drop, thermal, ESD and RF testing. | Mold tweaks and small BOM changes. Electrical design should be frozen. |
| PVT Production validation test | Can the factory build it repeatably, at yield, with the real line and fixtures? | Units built on the production line with production fixtures and test software. Many are sellable. | Process fixes only. A design change here sends you back to DVT. |
The cost of a change rises sharply across these builds. In EVT, moving a connector is a layout edit and a new board order. In DVT, it may mean re-cutting a steel mold. In PVT, it also means re-qualifying fixtures, updating test software and possibly retesting for certification.
Why does Duobay add gates before EVT?
Many failed hardware projects lose money before EVT, because the team commits to a custom board before it knows whether the chip, battery and radio can meet the product's needs. We add two gates in front.
- Gate 0, Feasibility (1–2 weeks). Requirements, chip and module selection, a costed BOM with a landed DDP estimate, the certification path and a risk register. No hardware is built. What Feasibility covers.
- Gate 1, Works-like prototype (3–6 weeks). Proven modules on a custom carrier board. Your model runs on the target silicon, and we measure power draw, latency and RF numbers before anyone draws a custom schematic.
For an AI device, Gate 1 is where software teams learn the most. You find out whether wake-word detection fits in the microcontroller's RAM, whether audio upload over LTE-M drains the battery in hours instead of days, and whether on-device inference is fast enough. Those answers decide the custom board.
What do you receive at each gate?
This is the deliverable list we commit to, gate by gate. The same list sits on our How it works page and in each gate quote.
| Gate | Time · units | You receive | If you stop here |
|---|---|---|---|
| Gate 0 Feasibility | 1–2 weeks No hardware |
| Stop here and keep every document. Fee is credited if you continue. |
| Gate 1 Works-like prototype | 3–6 weeks 3–5 units |
| Stop and keep the units and all source files. |
| Gate 2 EVT · first custom PCB | 6–10 weeks 10–30 units |
| Stop and take the full design package to any factory. |
| Gate 3 DVT / PVT · small batch | 8–16 weeks 10–1,000 units |
| Move production to a factory of your choice with molds, fixtures and test procedures. |
| Gate 4 Mass production | Ongoing 1,000+ units |
| Transfer to another manufacturer with a documented handover. |
In our model, EVT is Gate 2, and DVT and PVT are combined in Gate 3. Combining them works for runs of 10–1,000 units, because the line is small enough to validate the process during the same build that proves the design. For volumes in the tens of thousands, PVT usually runs as its own build.
What should a software team check before signing off a gate?
A gate review should not be a demo. Ask for numbers measured on the units you received, and test them yourself.
At EVT
- Power budget against the estimate. Sleep current, active current and battery life under your real usage profile. A tracker that idles at 50 µA instead of 10 µA loses most of its battery life.
- Every peripheral brought up. Microphones, sensors, radio, charger, fuel gauge, LEDs and buttons, each with a pass or fail note.
- OTA and provisioning. A signed firmware update, a failed update and a rollback, all run on your hardware against your backend.
- A build you can reproduce. Your team builds the firmware from the repository on a clean machine and flashes a unit.
At DVT
- Reliability reports. Drop, temperature cycling, ESD (commonly tested to IEC 61000-4-2), button and connector cycling, and battery safety. Ask which failures occurred and how each was fixed.
- RF pre-scan margin. How many dB below the FCC limits the radiated emissions sit. A thin margin means a likely retest at the formal lab.
- Certification status. Which reports exist, who the grantee is, and whether the grant will be in your company's name.
At PVT
- First-pass yield and a list of each failure cause.
- Test coverage. What the end-of-line test checks on every unit: radio, audio path, current draw, firmware version.
- Traceability. Serial numbers linked to IMEI or MAC, firmware version and test results.
- A locked BOM with second sources for parts with long lead times.
When can you walk away, and what does it cost?
You can stop at any gate. The real question is what you hold when you stop, and what it takes to restart somewhere else.
| Stop after | What you hold | What restarting elsewhere takes |
|---|---|---|
| Gate 0 | Architecture, costed BOM, certification plan, risk register | Another team can quote directly from the documents. |
| Gate 1 | Prototype units, schematics, layout source, firmware source, measured numbers | A new team reviews the design and starts the custom board. |
| Gate 2 (EVT) | Full electrical, mechanical and firmware package, EVT test report | Any capable factory can build DVT from the package. Budget a few weeks for their own design review. |
| Gate 3 (DVT/PVT) | Molds, fixtures, test procedures, certification grants in your name, locked BOM | Ship molds and fixtures to the new site, run a trial build, re-verify yield. Certification usually carries over if the design is unchanged. |
Two things make an exit work in practice. First, the files must be handed over at each gate, while you still have leverage. Second, the contract must assign ownership when each gate is paid. Our IP ownership page explains how we do both, and the article on what belongs in a source-file handover lists what to check.
What usually goes wrong at each stage?
These are common failure patterns across the industry. Planning for them early is cheaper than discovering them at the gate review.
- EVT: antenna performance drops once the board sits inside the real enclosure, near a battery and a hand. Budget for antenna matching on the first spin.
- EVT: the battery chosen in Feasibility cannot deliver LTE transmit peaks at low temperature, which causes brownout resets.
- DVT: the formal lab finds an emission that the pre-scan missed, often from a switching regulator or a display cable.
- DVT: a part in the BOM goes on allocation or end of life. This is why we lock second sources before production.
- PVT: yield is fine at the start of the run and falls later, because a fixture wears or a reel of parts comes from a different lot.
How long does the whole sequence take?
Using our gate ranges: Feasibility takes 1–2 weeks, the works-like prototype 3–6 weeks, EVT 6–10 weeks, and DVT/PVT with certification 8–16 weeks. A second EVT spin, a mold change or a certification retest adds time. Certification lab queues vary. We plan the lab choice in Gate 0 and run formal testing at FCC-recognized labs outside mainland China. See how we plan FCC certification.
Questions about gates
Is a works-like prototype the same as EVT?
No. A works-like prototype (our Gate 1) uses proven modules on a simple carrier board so your software runs on the target silicon early. EVT (Gate 2) is the first fully custom PCB in the real form factor. How the prototype gates work.
Can we skip DVT and go straight to production?
You can combine DVT and PVT into one build for small runs, which is what our Gate 3 does. You cannot skip the reliability tests and formal certification. A device without an FCC authorization cannot be legally marketed in the US.
What happens to the molds if we leave after DVT?
Molds you paid for are yours and are labeled with your company name. You can ship them to another factory, though moving a steel mold takes freight time and a trial shot at the new molder. How ownership works.
How many units do we get at each gate?
Gate 1 ships 3–5 units, Gate 2 ships 10–30, and Gate 3 builds 10–1,000. Exact quantities are fixed in each gate's quote.
Do we pay for the whole project up front?
No. Each gate is quoted and paid on its own. Gate 0 Feasibility is a fixed-fee study, credited against later work if you continue. Engagement terms.
Want to see the full deliverable list and contract terms for each gate? Read How it works, then tell us about your device and we will propose which gate to start at.
Planning a device like this? Tell us what it needs to do. A Duobay engineer replies within 2 business days. Book a scoping call →