EVT, DVT, PVT Explained: What Each Build Must Prove

You do not go from a working prototype to 3,000 units in one step. You build in phases. Each phase asks a different question and makes more units under conditions closer to real production. In hardware, those phases are usually called EVT, DVT, and PVT, followed by MP (mass production).

Most explanations stop at definitions. This guide follows one fictional product through every phase, with real-looking numbers: Northline's NL-100, a fold-flat aluminum laptop stand with a friction hinge that must hold 1.2 to 1.6 N·m of torque. It also shows how the same idea works for formulated products like skincare, using Rainwell, a fictional serum brand.

The four phases at a glance

The key rule: write exit criteria before the build starts. If you write them afterward, you will write them to match what you got.

Give every build one question

Before each build, write down what it must answer, what counts as success, and what stops it. Northline's PVT plan:

Do not pack so many questions into one build that the result cannot explain what happened. If you change the hinge supplier, the anodize house, and the pad adhesive in the same build, a failure points everywhere.

Track yield from the first build

First-pass yield (FPY) is the share of units that pass every check the first time, with no rework. It is the best single signal of whether a process is under control.

Watch for the gap between first-pass yield and final yield. After rework, Northline's PVT shipped 493 of 500 good units, a final yield of 98.6%. The manufacturer's report led with 98.6%. Northline's dashboard showed 94.2%, because that is the number that says whether the process is in control. The difference is the "hidden factory" of rework nobody quoted.

What PVT found that DVT could not

At DVT, a technician set hinge preload by hand. Everything passed.

At PVT, 18 of the 29 rejects were hinge torque above 1.6 N·m. The data told the story: the first 100 units averaged 1.42 N·m and the last 100 averaged 1.55. All 18 rejects came in the second half of the build. Both hinge lots showed the same pattern, so the parts were not the cause.

The manufacturer checked the powered torque driver that sets preload against a calibrated reference. Over the build it had drifted to read 8 percent low, so each hinge was tightened harder than the display showed. The driver was not on the calibration schedule, because at DVT nobody used it.

The fix: the driver went on the calibration schedule, with a check against a reference at the start of every shift recorded on the traveler; 100 percent torque testing continued until capability was proven; and a corrective action was tracked to closure.

This is what PVT is for. The problem was invisible at 100 hand-built units. It appeared at 500 units at line rate. You want to find drift at 500 units, not at 3,000.

Formulated products: lab, pilot, commercial

Skincare, cosmetics, and other batch products use different names for the same logic.

Scale changes physics. Rainwell's pilot came in at 6,400 cP against a 5,000 cP limit, because the manufacturer's pilot homogenizer had far more shear than the founder's lab mixer. Three small sub-batches at different speeds found a setting that gave 4,200 cP, and that speed range went into the batch record. The founder also restarted the stability clock on a second pilot at the new setting, costing four weeks, because stability data on the wrong product proves nothing.

Plan the development, not just the builds

Both examples share a lesson. The equipment you get at scale is rarely the equipment you used at the bench. So some development is part of every transfer: fixtures, settings, sometimes a product change. Put it in the schedule and budget before you commit.

Northline held a $4,000 development reserve and three weeks of schedule for exactly this, and used two of the three weeks. Rainwell's extra pilot cost about $3,000 (illustrative) and four weeks.

After PVT: accepting production lots

Once production starts, each lot is usually accepted by sampling. Our guide to AQL sampling walks through Northline's first 1,000-unit lots.

Do not skip a phase

Skipping PVT or the pilot does not remove the problems. It moves them into your first commercial order, where they cost more and reach customers.

Frequently asked questions

What do EVT, DVT, and PVT stand for?

Engineering validation test, design validation test, and production validation test. EVT asks whether the design works, DVT asks whether it meets every spec when made the production way, and PVT asks whether the real line can build it at rate and yield.

How many units are in EVT, DVT, and PVT builds?

It depends on the product and risk. A common pattern for a simple hardware product is tens of units at EVT, around a hundred at DVT, and several hundred at PVT. Northline used 20, 100, and 500.

Can PVT units be sold?

Often, yes, if they pass every check and the process matches production. Northline planned to sell passing PVT units as launch stock and hold rejects for review. Agree this in writing, and make sure your minimum order quantity and pricing cover the PVT build.

Where does first article inspection fit?

First article inspection is a full measurement of the first parts made with production tooling and process, against every drawing requirement. It usually happens at the start of PVT. It proves the process can make one part to the drawing. PVT yield and validation prove it can repeat.

Next steps

1. Check that your build plan and critical features are defined with the free manufacturer readiness check.
2. See what first-pass yield below target does to your margin in the product cost calculator.
3. The Contract Manufacturing Playbook covers transfer, scale-up, and validation in full, with a Tech-Transfer Plan, a First Article Inspection Report template, and a Validation Plan & Protocol, each blank and filled in for Northline.