Skip to content
HardwareMap
esc
  • PageBrowse the directory219 parts
  • PageHow the index workshubs · dates · rules
  • PageSubmit a startupget a part number
  • PageApplication noteslists · build guides
  • CategorySilicon & ComputeSIL
  • CategoryIndustrialIND
  • CategoryDefenseDEF
  • CategorySpace & AerospaceAER
HardwareMap application noteAN-043Rev 26.09Build guides

AN-043 Build guides · cluster

Prototype to Production: The Stages Nobody Draws Correctly

Prototype to production in four gated builds: EVT, DVT, PVT and ramp. What each proves, unit counts, exit criteria and where schedules slip.

by the HardwareMap editors4 min read

Contents
  1. 1Prototype to production in one table
  2. 2EVT: does the design work?
  3. 3DVT: does it survive?
  4. 4PVT: can the line build it?
  5. 5Where the stages actually overlap
  6. 6Frequently asked questions

Prototype to production runs through four gated builds after your last prototype: EVT proves the design works with production-intent parts, DVT proves it survives testing on tooled parts, PVT proves the factory builds it at rate and yield, then ramp. This is stages five to seven of how to start a hardware company, and each gate is a pass or fail.

Most diagrams draw these as four tidy boxes in a row. In practice they overlap, loop back and run in parallel with tooling and certification. This note draws them the way they actually run, for a connected consumer device. Industrial and medical products follow the same order with longer DVT.

Prototype to production in one table

BuildQuestion it answersParts come fromTypical unitsExit when
EVTDoes the design work?Printed, machined, soft tooling; first custom boards10 to 50All functions pass on the bench, BOM is stable
DVTDoes it survive and certify?Production mold (T1 shots), production boards50 to 300Reliability tests pass, certification pre-scans pass
PVTCan the line build it?Production parts, production line, production fixtures100 to 1,000+Yield and cycle time hit targets on the real line
RampCan we keep building it?ProductionYour forecastCost, yield and return rate stable over several lots

Unit counts are typical for a consumer device and scale with how many units each test destroys. A drop and thermal test plan can consume 30 to 60 DVT units on its own.

Ramp gets no box of its own on most diagrams, and it should. The first production lots are where component substitutions, operator turnover and a second shift show up. Keep the engineering team on the factory floor, or on a daily call with it, until three or four lots in a row hold the PVT yield. Count field returns by failure mode from the first shipment, because the return rate is the number that tells you whether DVT tested the right things.

EVT: does the design work?

EVT is the first build where the parts are meant to look like production, even if they come from a print farm or a machine shop. The circuit board is the production layout, on production stack-up, assembled on a real line rather than by hand. The firmware runs the full feature set.

What EVT tests:

  • Every function against the product requirements document, on every unit.
  • Power: battery life, charge time, sleep current.
  • Fit: do the parts assemble without force, with the clearances in the drawings?
  • Thermal: hot spots under worst-case load.
  • Early emissions pre-scan, so a radio or clock problem shows up before the board is frozen.

EVT usually produces one or two board spins and a list of mechanical changes. That is the point. The cheapest place to change a design is before steel is cut. Every change made here should follow the design for manufacturing rules, or it will come back at DVT.

DVT: does it survive?

DVT units come from production tooling. That matters because a printed housing does not have the same wall thickness, shrink, parting line or knit lines as a molded one. A snap fit that worked in SLA resin can crack in polycarbonate.

DVT runs the reliability plan:

  • Drop and shock at the heights and orientations in your spec.
  • Thermal cycling and humidity, to find solder joint and seal failures.
  • Life cycle on buttons, hinges, connectors and anything that moves.
  • ESD on every touchable surface.
  • Certification pre-scans for emissions and safety. In the US, unintentional and intentional radiators fall under FCC Part 15. Formal lab testing is often run on DVT units.

Tooling changes found at DVT are the expensive ones. Ask the mold maker to cut critical dimensions steel safe, so the fix removes metal from the mold. Adding metal back means welding and re-cutting. Texture and polish go on after dimensions are approved, not before.

For aerospace and defense parts, first article inspection is formalized: SAE AS9102 defines the report that proves a production part matches the drawing. Consumer products borrow the idea without the paperwork.

PVT: can the line build it?

PVT is a build of the factory, not the product. The design is frozen. The question is whether operators, fixtures and test stations can produce good units at the planned rate.

What PVT measures:

  • First-pass yield at each test station and end of line.
  • Cycle time per station, and whether the line is balanced.
  • Fixture repeatability: does the test fixture pass good units and fail bad ones?
  • Packaging and labeling, including drop tests on the packed unit.
  • Workmanship against an agreed standard. For board assembly, most contracts reference IPC acceptability classes.

PVT ends with a golden sample signed by both sides, a written yield target and acceptance criteria. Choosing the right partner for this stage is covered in choosing a contract manufacturer.

Where the stages actually overlap

The tidy diagram hides four loops that set the real schedule:

  1. Tooling kicks off during EVT, not after. Mold lead time often exceeds the EVT build itself, so teams release long-lead parts first and accept some risk.
  2. Certification runs through DVT and into PVT. Lab queues are booked weeks ahead. A failed scan sends you back to a board spin.
  3. Component lead times run in parallel from day one. A part with a long lead time has to be ordered before the design is frozen, which means forecasting before you know the yield.
  4. Failed builds repeat. An EVT2 or DVT2 is normal on a first product. Plan one in the schedule, and you will usually be on time.

The cost of each build, and how it fits the 18-month budget, is in hardware startup costs.

Frequently asked questions

What do EVT, DVT and PVT stand for?

EVT is engineering validation test, DVT is design validation test and PVT is production validation test. EVT checks that the design works using production-intent parts. DVT checks that it survives reliability and certification testing using parts from production tooling. PVT checks that the factory can build it at the planned rate and yield. Mass production starts after PVT passes.

How long does it take to go from prototype to production?

For a connected consumer device with one or two molded parts, EVT through PVT usually takes six to twelve months. Tooling lead time, certification lab queues and component lead times set the floor. Each failed build that forces a repeat adds one to three months. Products with safety, medical or aerospace qualification take longer because qualification testing sits inside DVT.

Can a startup skip EVT or DVT?

It can merge them on a simple product, such as a board in an off-the-shelf enclosure, but it cannot skip the tests. Skipping a build moves its failures to the next one, or into customers' hands. The usual compromise is a combined EVT and DVT build with production tooling cut early, accepting the risk of a mold change if a test fails.

When should injection mold tooling be cut?

Usually after EVT, once the mechanical design has passed fit and function checks with printed or machined parts. DVT units should come from the production mold, because DVT tests the real material, wall thickness and parting line. Leave steel safe on critical dimensions so late fixes remove metal from the mold rather than adding it, which is slower and costlier.

What yield should a hardware product hit at PVT?

There is no universal number, but many consumer electronics teams set a first-pass yield target in the mid to high nineties at end-of-line test before calling PVT complete. Lower yields mean rework cost per unit and usually point to a design or fixture problem. Agree the target and the test limits with the contract manufacturer in writing before the PVT build starts.

HardwareMap tracks the companies that made it through PVT and into production. Building one? Submit it and get a part number.

Building one?

Submit it to the index and get a part number. Every entry is reviewed by hand.

Submit a startup

Related notesAll notes →

Sponsor spot20/20 open

Put your startup in front of everyone browsing the directory of 219 hardware companies.

$50/ week

First 10 spots · 10 left

$70/ week

Next 10 spots

Book through @icantcodefyi on X.