New Product Introduction
EVT, DVT, PVT: What Each Build Phase Requires From Your Supplier
The short answer
EVT, DVT and PVT are the three validation builds a hardware product goes through between a working prototype and mass production. EVT asks whether the design works. DVT asks whether it meets specification reliably in the real materials. PVT asks whether the factory can build it at volume and yield without your engineers in the room.
The three phases at a glance
| Dimension | EVT | DVT | PVT |
|---|---|---|---|
| Question answered | Does the design work? | Does it meet spec and survive? | Can we build it at volume? |
| Typical quantity | 5 to 50 | 50 to 300 | 100 to 1,000+ |
| Parts made by | Machining, 3D printing, hand fab | Soft tooling, near-final process | Production tooling and process |
| Tolerances | Loose, nominal is fine | Drawing tolerances enforced | Enforced, plus capability data |
| Documentation | A STEP file is enough | Real drawings, GD&T, FAI | Control plan, yield data, inspection records |
| Certification testing | No | Yes: FCC, CE, UL, reliability | Confirmatory only |
| Typical lead time | 1 to 3 weeks | 4 to 10 weeks | 6 to 14 weeks |
| What kills you here | Design does not function | Tooling reveals it is unmanufacturable | Yield is 60% and nobody knows why |
EVT: prove the concept functions
At EVT you are answering whether the mechanism actuates, whether the board fits the enclosure, whether the thermal design holds, and whether the assembly goes together in a sane order. You are not answering questions about cost or appearance yet.
Ask for machining or 3D printing, no tooling. Tooling at this stage is money set on fire, because the design will change. Prioritize turnaround over price: a one-week part at three times the cost is cheaper than a three-week part, because your engineers are idle in the gap.
The most common EVT mistake is cutting a soft tool to save time later, then changing the design twice and cutting it again.
DVT: prove it meets specification
DVT is where hardware programs actually get decided. It is the first time the part is made the way it will really be made, and the first time the design meets physics it was insulated from at EVT.
You need real drawings with real tolerances, not just a model, because now somebody has to inspect against them. You need a first article on the first parts off the tool. You need cosmetic approval with a signed master sample. And you need certification units, which is why DVT quantities are higher than they look: labs consume units and destructive testing consumes more.
Settle tool ownership in writing before the purchase order. Who owns the tool if you leave, what its expected life is, and who pays for engineering changes to it.
PVT: prove the process, not the parts
PVT parts come off production tooling, on the production line, run by the operators who will actually run it. The thing under test is not really the parts. It is the process.
Ask for a yield report with a defect Pareto rather than a pass or fail. Ask for capability data on critical characteristics. Ask for the control plan and work instructions. And confirm the operators running PVT are the operators who will run production.
The classic PVT mistake is running it with your best technician. If your most skilled person hand-builds the units, you have validated that person, not the process.
Cost per unit does not fall smoothly
| Phase | Unit cost driver | Direction |
|---|---|---|
| EVT | No tooling, expensive per-part processes | Moderate |
| DVT | Tooling amortized over a small run | Highest per unit |
| PVT | Production tooling over larger volume | Falling |
| Mass production | Tooling amortized, line optimized | Lowest |
A DVT part often costs more than the EVT part it replaced, because you just paid for a tool and divided it across 150 units. Budget for it.
What to put in the RFQ at each phase
The same part gets sourced very differently depending on which build it is for. Telling a supplier which phase you are in changes what they optimize for, and most buyers never mention it.
| Dimension | EVT | DVT | PVT |
|---|---|---|---|
| What to optimize for | Speed above all | Tooling quality and DFM | Yield, documentation, repeatability |
| What to send | STEP file | Dimensioned drawing with GD&T | Full drawing package and spec |
| What to ask for | Fastest possible turnaround | Written DFM feedback before tooling | Yield report and capability data |
| Inspection to request | None, or a spot check | First article on first parts off tool | FAI plus in-process controls |
| What to negotiate | Nothing, just buy speed | Tool ownership and change costs | Price at volume, capacity commitment |
The three sourcing mistakes that cost the most
Switching suppliers between DVT and PVT to save money. The new shop inherits a tool it did not cut, a process it did not develop and a drawing package it did not review. Whatever you saved on unit price, you tend to spend on requalification, and you lose the schedule too.
Skipping DFM review before tooling. This is the expensive one. A design change before steel is cut is a drawing revision. The same change after is a new tool.
Sourcing each line item separately. A product is a bill of materials, not a part. Teams that source forty line items through twelve vendors spend their engineering time chasing instead of designing, and every mismatch in lead time becomes a schedule slip. Consolidating the sourcing is usually worth more than the unit-price savings that fragmenting it produced.
How long each phase takes, and where the time actually goes
Schedules slip in predictable places. Knowing which parts of each phase are compressible and which are not is the difference between a plan and a wish.
| Phase | Typical duration | The part you cannot compress |
|---|---|---|
| EVT | 3 to 6 weeks | Your own design iteration between builds |
| DVT | 8 to 16 weeks | Tooling lead time, then certification lab queues |
| PVT | 6 to 12 weeks | Running enough units to get a meaningful yield number |
| Between phases | 2 to 6 weeks each | Reviewing results and deciding what changes |
Durations assume the design is stable entering the phase. They do not survive a mid-phase redesign, which is the most common cause of a slipped hardware schedule.
Compressing the schedule without breaking the process
There are legitimate ways to pull a hardware schedule in, and there are ways that only move the delay somewhere less visible.
What genuinely works: ordering long-lead items before the design that surrounds them is finished, because a twelve-week connector does not care that your enclosure is still changing. Booking certification lab slots before you have units, since the queue is often longer than the testing. Running DFM review during EVT rather than at the start of DVT, which is free and routinely skipped. Building EVT and DVT quantities in the same order where the process allows, to avoid a second setup charge. And sourcing the whole bill of materials through one point of contact instead of twelve, because chasing is the hidden time sink in every hardware program.
What does not work: skipping the DFM review, switching suppliers to save a week, cutting production tooling before the design is frozen, and reducing PVT quantity below the point where the yield number means anything. Each of these buys a few weeks now and costs more later, usually at the worst possible moment.