How Many Prototype Units You Need for Testing

One Unit Is Almost Never Enough

Most first-time developers budget for a single prototype, and most of them lose two to four weeks because of it. The reasons are mundane. Destructive tests consume the unit. A drop test damages the housing you also need for user sessions. The unit is sitting in a certification lab while your investor meeting is on Thursday. Someone opens it to probe a signal and cracks a boss.

The right question is not "how many prototypes should I make." It is "what tests am I running, in what order, and can they share hardware?" Count from the test plan. The number falls out of it.

Count by Test, Not by Feeling

Write every planned activity in one column and mark whether it consumes, damages, blocks, or leaves the unit intact.

• Consuming tests. Drop to failure, burst pressure, tensile pull on a joint, overload of a mechanism, thermal runaway. Each one costs a unit permanently. Testing to failure is the point, and the results are useless if you stop early to protect the sample. Plan the sample count the same way you would for any drop-test program: multiple orientations, multiple samples per orientation.
• Damaging tests. Environmental soak, UV exposure, salt spray, water ingress. The unit survives but is no longer representative for cosmetic or user testing.
• Blocking tests. EMC pre-scans, safety evaluation, biocompatibility. The unit physically leaves your building for one to six weeks. If your only unit is at the lab, everything else stops.
• Non-consuming activities. Photography, investor demos, internal review, most software development. These can share, but not with each other on the same day.

Now add a parallelism column. Two workstreams that need hardware in the same week need two units, regardless of whether either test destroys anything. That single insight usually moves the count from one to three.

Numbers That Show Up in Practice

Simple mechanical consumer product (no electronics, molded or printed housing): three to five units. One for fit and finish, two for destructive testing, one or two spares for user sessions.

Electronic product with an enclosure: five to ten. Firmware developers need one each, and they need it for months. Hardware debug needs a board with test points exposed. Certification pre-scan takes one. Demo and photography take one that stays pristine.

Product going to a formal user study: the study size sets it. Sequential one-on-one sessions can run on two units, one in use and one charging or being reset. Simultaneous sessions or a multi-week take-home study needs one per participant, which is why a twelve-person diary study is a real budget line. The planning tradeoffs are covered in user testing with a prototype.

Medical or safety-regulated device: ten to thirty. Biocompatibility, sterilization validation, electrical safety, and usability all consume units, and each test typically requires multiple samples from the same build to demonstrate consistency.

Field trial: five to twenty-five, depending on how many sites and how long. Units leaving the building do not come back on schedule, and some do not come back at all. Build in attrition, and read field testing a prototype before you set the number.

Where the Marginal Unit Is Cheap

The cost curve is not linear, and knowing its shape lets you buy more units for the same money.

Cheap to duplicate: 3D printed parts, where the second unit costs roughly the same as the first minus setup; PCB fabrication, where five boards often cost barely more than two because the panel and setup dominate; and urethane cast parts, where a silicone tool amortizes over fifteen to twenty pulls. That last one is the classic move for quantities in the ten to fifty range, described in urethane casting for low-volume production.

Expensive to duplicate: CNC machined parts, which are almost linear in quantity because machine time repeats; hand assembly and wiring, which is pure labor per unit; hand-painted cosmetic finishing, which can be several hundred dollars per housing; and purchased components with high unit cost such as sensors, displays, and battery packs.

The practical consequence: when the housing is printed, ask for five instead of two, because the delta is small. When the housing is machined from a solid billet, ask hard whether the second one is really needed. The full breakdown of what drives these numbers is in how much it costs to make a prototype.

Not Every Unit Has to Be the Same Unit

The most common budget mistake is building five identical full-fidelity prototypes. You almost never need that. Build a mixed fleet:

• One hero unit. Full cosmetic finish, final color, correct graphics. It never gets opened and never goes to a lab.
• Two or three working units. Unpainted, visible fasteners, oversized housing if that makes assembly easier. These carry the real functionality.
• One or two structural mules. Correct geometry and material, no electronics, built purely to be destroyed.
• Bench rigs. A development board, a motor, and a power supply on a plate, with no enclosure at all. Firmware and control work does not need a housing, and giving the software team a bench rig frees a real unit for something else.

A mixed fleet of six often costs less than three identical full-fidelity units and answers more questions. The same logic underlies most of the advice in how to cut prototype costs.

Deciding the Number

A workable procedure: list the tests, mark consumption, mark parallel workstreams, sum the units, then add one spare for every three units in the plan. Round up rather than down, because a re-order mid-program costs the setup charge again plus one to three weeks of lead time, which is nearly always worse than having bought an extra unit up front.

Then hold the quantity against the build stage. Early concept builds should be few and fast, because the design will change. The quantity ramps at the verification stage, when the design is stable enough that the results still apply after the next revision. Testing a large fleet of a design you already know is changing is the other way to waste the budget, and it is closely related to the failure patterns in reliability testing for a new product.

Sizing Your Build

Projects House scopes prototype builds against a written test plan, so the quantity, the fidelity mix, and the fabrication method are chosen per unit rather than repeated blindly. Send your product concept and what you need to prove through our contact form.