# EVT, DVT and PVT: build stages for small hardware teams

> What EVT, DVT and PVT builds are for, how many units each needs and what should gate the next stage on a small hardware product.

- Author: Milo Perović, Founder & lead engineer
- Published: 2026-11-02
- Topic: Process

Most hardware schedules we are shown jump straight from "working prototype" to "production". In between sit three build stages that larger companies call EVT, DVT and PVT, and skipping them is where small teams lose the most time and money. This post is for founders and product managers planning their first manufactured product who want to know what each stage is for, how many units it needs and what should block the next one.

You do not need a corporate stage-gate process to benefit. You need to know which question each build answers, so you stop asking a pile of 3D-printed prototypes to prove things only tooled parts can prove.

## What EVT, DVT and PVT actually answer

Each stage is a build of real units with a single, narrow purpose. The names are industry shorthand, and the boundaries blur on small projects, but the questions stay the same.

- **EVT (Engineering Validation Test):** does the design work? Electronics, firmware and mechanics are proven against the requirements, usually on near-final PCBs inside printed or machined enclosures.

- **DVT (Design Validation Test):** does the design survive reality? Units are built with production-intent parts, including first samples from hard tooling, and go through environmental, drop, EMC and radio testing.

- **PVT (Production Validation Test):** can the factory build it repeatedly? Units come off the real line, at the real rate, with the real test fixtures and operators.

The trap is treating these as quality levels rather than questions. A beautiful EVT unit tells you nothing about mould shrinkage, and a perfect PVT run cannot fix an antenna that was never validated.

## Before EVT: prototypes that are allowed to be ugly

Ahead of EVT there is usually a looser phase of breadboards, development kits and first-spin PCBs. Its job is to retire the big technical risks: will the sensor resolve what we need, does the radio reach, does the battery last long enough on paper and on the bench.

We keep this phase deliberately cheap. Boards are hand-assembled or ordered in single digits, enclosures are printed overnight, and firmware is just enough to exercise the hardware. The output is not a product but a list of decisions you are now confident about, which is exactly what the Abstract phase of our [CAEC process on a typical project](https://miloperovic.com/journal/what-the-caec-process-looks-like-on-a-typical-project/) is built to produce.

## EVT: proving the engineering

EVT units should be as close to the final electronics as you can make them: the intended microcontroller, the intended radio module, the final power path and connectors. Mechanics are usually SLS, SLA or CNC parts that match the CAD, not the eventual tooled plastic.

Typical EVT checks include:

- Every requirement tested on the bench, with pass, fail or "not testable yet" recorded against it.

- Power budget measured in each operating mode, not estimated.

- Radio range and connection stability in the real enclosure, since a printed housing in a similar material is a far better proxy than a bare board.

- Firmware feature-complete enough to run long soak tests, including OTA updates and recovery from brown-outs.

- First fit and assembly checks: cable routing, screw bosses, button travel, light pipes.

For a small product, 5 to 20 EVT units is usually enough. You want enough to run tests in parallel and to see unit-to-unit variation, not so many that a board respin leaves a drawer of scrap. Expect at least one PCB revision out of EVT; planning for it is cheaper than hoping.

## DVT: the stage small teams skip

DVT is where the design is frozen and then attacked. Parts should be production-intent: the final PCB revision from the intended assembler, and enclosure parts from the first shots of the hard tool (often called T1 samples) rather than prints. Material, colour, texture and wall thickness all change behaviour, and only tooled parts show sink marks, warpage, snap-fit stiffness and real gasket compression.

This is also where formal or pre-compliance testing belongs. EMC, radio and safety testing on DVT units means the results apply to what you will ship. Testing EVT hardware is useful for finding problems early, but a certification report on a design that then changes is a report you may need to repeat.

DVT quantities for small products often land somewhere between 20 and 100 units, depending on how many are consumed by destructive tests. Drop, thermal cycling, ingress and long-term ageing tests eat units, and you want a few left over for field trials with real users.

### What usually comes out of DVT

Mould adjustments are the most common finding: a boss that sinks, a clip that is too stiff, a gap that is too visible. These are steel-safe changes if the tool was designed with that in mind, and expensive ones if it was not. Electronics changes at this point should be minor, such as component value tweaks or a filter added after an EMC scan, not new features.

## PVT: proving the factory, not the design

By PVT nothing about the design should be in question. The goal is to prove the process: line layout, work instructions, fixtures, firmware programming, labelling, packaging and yield. Units are built at or near the intended rate by the operators who will build production.

The most important artefact here is your end-of-line test. If it cannot catch a missing component, a weak radio or an uncalibrated sensor quickly and reliably, you will find out from customers instead. We covered how to design that test in [end-of-line testing for your first production batch](https://miloperovic.com/journal/end-of-line-testing-for-your-first-production-batch/).

PVT units are often sellable if they pass, which softens the cost. Quantities depend entirely on the factory and the product, from a hundred or so units for a niche device to a meaningful fraction of the first production order.

## Gates between stages

A stage is finished when its question is answered, not when the calendar says so. We write the exit criteria before the build starts, so nobody argues about them afterwards.

GateTypical exit criteria

EVT to DVTAll requirements tested; no open blocking bugs; PCB and CAD frozen; tooling released

DVT to PVTEnvironmental, drop and pre-compliance passed; mould changes signed off; certification under way or complete

PVT to productionTarget yield reached; end-of-line test stable; packaging and labelling approved; certification in hand

Small teams can merge stages, and sometimes should. A product with an off-the-shelf enclosure has no tooling risk, so EVT and DVT can be one build. A very low-volume product built by hand may treat PVT as the first production batch with extra inspection. What you should not do is merge stages to save time while the risks they cover are still open.

## What we would do on a small budget

If you are planning your first product with EVT, DVT and PVT in mind, these are the decisions that matter most:

- Write the question each build must answer, and its exit criteria, before ordering parts.

- Budget for at least one PCB respin after EVT and one round of mould adjustments after DVT.

- Do not start hard tooling until EVT has closed, and do not book certification until DVT parts exist.

- Count destructive tests when sizing the DVT build, then add spares for field trials.

- Design the end-of-line test during DVT so PVT validates it rather than invents it.

- Merge stages only when you can name the risk that no longer applies.

None of this needs a large organisation. It needs a clear view of what each build is for, and the discipline not to ask a prototype to answer a production question.
